Internet Engineering Task Force Interim Meetings, 8 Feb 2023

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: IETF-CBOR-20230208-1500

Description

CBOR meeting session at IETF
2023/02/08 1500

https://datatracker.ietf.org/meeting//proceedings/

A

um Hello, the two of you, um even though I would generally try to start minutes on top of the hour. um I think it would make sense to wait a few minutes, because there's three of us and while that is a group um I think we will have a more productive interaction if there are a few more.

A

Interesting, a lot.

B

Yeah Marco responded to me on the jabber before mid Echo started, so my response on zulub actually is now part of the meeting, because it is in the the time span of 10 minutes before the meeting yeah yeah. So I just wondered why there is no meeting materials.

B

A link on the data tracker upcoming uh Paige and Marcus Lewis thought that this is because we didn't submit a separate agenda.

A

Oh, never look.

A

Yeah we used we used to have uh we used to have a dummy agenda, but given that the actual actual agenda is always in the minutes anyway, yeah.

B

Yeah um so I think we should always submit the single line agenda that has a pointer to the notes, page yeah.

A

I used I used to do this um I think I should yeah. I should probably should add this retroactively for the for the ones in this term which Barry had set up um yep we'll do that.

C

Yeah: okay, I, don't.

D

C

Intentional, maybe it's something to discuss in the tools team.

C

B

D

C

Material page exists anyways, so why not link into it? Even though there's no agenda, you may still have something else.

B

Yeah, so it's probably to punish meetings that didn't update.

B

A

I know that we will be a rather short group, given that Ira said that he won't be here today and Barry won't be here either.

B

um Yeah I, don't know where the Michael is coming.

A

B

Feedback on the mailing list, so I'm not at all unhappy.

A

Let's, let's use those two minutes to go through the formalities, so this is a sibo meeting. um We are an ITF meeting, a place, be nice to each other and follow the note well and I yeah.

D

A

Length of the minutes is um I'll post here again, just in case anyone listened and thanks Marco for adding to the minute.

A

We have one basically two items on the agenda today: cddl 2.0 and zero path, both of which we have in your slides customers. So maybe it's just maybe you could already just start them up and if Michael still joins.

A

Also give brand.

A

We don't we don't have to hit off into the into the most important parts right away. Yeah.

B

So let me just uh do the silver path thing on on one slide, um so this popped up uh more or less independently and uh I. Think it's a pretty obvious thing. I mean I can probably reveal the secret that that I actually became an editor of the Json past action when, with the view to actually being able to use the work in in the Civil uh context.

B

uh So this was uh quite foreseeable and now somebody has stepped up and actually started uh defining what they think and civil path is and um I. Think it's pretty interesting.

B

So my initial view was that a siwa adaptation of Json pass would essentially work with Json path, as is but then we don't get access to the more extensive generic data model of sibo, so we don't get access to bytes or to map keys that are not text strings.

B

So that's one thing, and the other thing of course is that sibo has a third container uh kind, which is tag so we we Jason, has arrays and and Maps, which are called objects in Json and sibo has arrays maps are a little bit extended because the keys can be anything not just strings and then we have tags. So those are the obvious places where adjacent zero path would extend uh Json path, um so I I always thought, let's finish Json path and we are almost there.

B

So yesterday we had a an interim where we essentially got stuck on the type system issue that we probably need to cover more days to understand, uh but we should then go into our Google Plus call.

B

um So, uh interestingly, uh the other Michael has, with the long last name that I kind of remember has generated a rust implementation. uh Instead of writing the document. First, he wrote the code first and I think that's that's not at all a bad way. To uh start with that, and one interesting thing he contributed was: let's not use text for this, but structured sibo.

B

So we we are not going to make the detour over a URI to a CI, but we as we will be starting with a CRI in the first place, whether this is not a UI, but you understand what I mean so.

A

So not quite so, the the um the format of the seawall path would be sibo already. Is that yes,.

A

Which, of course, could be shown in diagnostic notation, exactly.

B

So um the diagnostic mutation uh that that would result from Michael's proposal here um is uh not that bad, so um I I think we could live with that. uh But I I mostly wanted to raise this. So so people who know that we will have to do that and obviously the The Next Step um somewhere in the cdbl world, is doing what dsdl did by adding schematron to relax Ng.

B

uh We will probably need some way of attaching predicates uh to places in our data models and something like zero pass or an extended form of sibo path. Maybe, the the obvious way to to do this so yeah that's going to come up- maybe not this week, maybe not this year, but definitely civil path is something that we want to have in in our tool set.

B

So please have a look at this GitHub issue, which is on the seabor GitHub IO page.

B

um We always have a hard time putting up issues for a working group, but I think that that's pretty much as good as it gets um and look at the issue like look at the linked rust implementation. I didn't put the zinc on the slide because you can get through it from from there, and it would be great to have some some discussion that that might include the sibo meeting with everything so that that's all I have about this. One.

A

Thanks um I personally haven't had a chance to look into it, but sounds sounds like something that would be good to have.

B

Yeah, so it's probably most useful for those people who are in in the Venn diagram between uh the the Json path, people and the receiver. People, and probably that part of the Venn diagram is pretty small at the moment and I'm trying to get more people interested.

A

I I was more thinking. I'm I was not so much thinking about Jason's path, but more in the direction of like people. People don't use the traditional tools, set, cut, orc, Etc so much anymore, but start using um JQ and I think this would kind of having something that works with c-bob with seabor and I. Think sibo path would fit. The bill here um might might make things good better. Here.

B

Yeah so maybe at some point we can have a discussion about tools that might uh like uh civil support added to them. um So JQ certainly is a nice uh tool. I found that it's very different uh in usage from what I would use a CQ for.

B

um So I I wasn't so motivated to to actually uh get a student to submit a patch to to JQ. uh But let's see uh so.

A

Good I I was just thinking more on the high level, not not necessarily like exactly you see JQ, but it might. It might find a place into link yeah.

A

Yeah any questions on that topic, Marco.

D

Specific for me either thanks.

A

um Being 10 past I I'm afraid we won't have Michael here today, so um I'd say: let's go to onto cdl2 anyway and hope to produce something something useful. Also in the minutes for those who review this later.

B

Yeah and we also have the recording so um yeah- let's talk about the the current bike, so I just wanted to to implement something, so we can play with it. So this is not necessarily exactly what we will put into a CDL 2.0. On the other hand, um I find what we came up with so far, actually quite useful and and it's almost an 80 solution to the the things I actually need. So, uh let's see, but um essentially it's it's trying to uh address uh three requirements.

B

uh One is a requirement for someone who has a slightly larger cdda project to construct uh the cddl from multiple files, so files are taking the role of a module here. I think we all have tried uh languages that do multiple modules per file and that that can be um how to use so maybe equating files and modules for the moment I get to it.

B

This an individual, more detail later uh works, so when I I'm doing a project where I need to do A4, I may want to have multiple files and I may want to integrate them in a very defined way. Of course, I can do this with the make file and Cat, um but yeah that it's more interesting uh to to actually have a file that that has a standardized meaning that every tool can understand or that at least every tool can can somehow process.

B

So that's one thing, uh maybe more important is to be able to reference existing cddl as a library.

B

um So what what the previous one is handled by the include director, if this is handled by the import directive, which otherwise are very similar, but the difference here is that when you use an existing CDL file as a library, you do not necessarily need everything. That's in there. So when I write Syria specifically for a project, then of course what I write there is meant for the project. So it is all meant to go into that project.

B

But if I use an existing city as a library, then then maybe I only need cozy, encrypting and not course designer or vice versa, so that that's what import is doing. And uh finally, there is a very small appendix here that actually does something that reminds uh of namespaces so that that's the as clause in an include or an import.

B

So that's really what this little spike is trying to do. um We still have a cdl2 plan document, which has a few other things that also have to go into a CDL too. uh These are, of course, are not going away, but this is I, think the the most useful start for uh doing some, some module structure and some some namespacing foreign.

B

So um the approach here is uh pretty weird, um so the idea is that those things can all be done by a cddl processor, so I'm not calling it cddl preprocessor, because the cddl processor is fully aware of the semantics of of what it's actually throwing around here. So it's not like you see preprocessor, it doesn't understand. C, the cddl processor actually fully understands what a rule is and and how you, for instance, what substitute names in that and and uh and so on.

B

um So the the CDL processor would make use of the fact that we planned cddl2 to be a superset of cdl1 and what essentially process uh everything that doesn't work everywhere, but in almost all cases what process everything that is not the video one um and turn it into compatible cda1.

B

So we can continue to use cd-day, Cogen and and Andrew wise CDD air tool and the early cddl tool, which I hope will go away within this year um and um just make sure that the studio2 introduction is not disruptive. Nobody feels like they cannot use cddl tool because their tools are not yet supporting them.

B

The other effect of having the serial processor spit out cdl1 is that whatever processing is going on is very inspectable. So you, you, never have a question mark in your mind of what's actually going to go on, you just look in the output of the cddl processor and we'll be able to to find out what what your directives actually did.

B

So this is an implementation uh approach. It's, of course not the only implementation approach, not the only valid one, so some implementations will simply integrate that and actually that's going to happen in the inverse Direction with cdlc, which is a cddl processor and we'll get all the other parts of the early cddl tool in in time. But on the other hand, if you have an implementation that that really is trying to be simple and focused on a specific issue like code generation, uh maybe you don't want to implement all of of City there.

B

One even and uh CDL processor might be able to help you so cdlc has a prototype, the genic genericizer.

B

So if your cddl tool doesn't support generics, that's not a problem, because you can convert your your CDL input to a generics, free, cddl and process that with your tool, so that that may not be satisfying for another reason, because maybe the way the the studio one is generated is not what your tool needs, but that is of course, a separate issue.

B

So um the the Prototype right now is in the cdsc tool, which is essentially a converter or compiler for cddl, and um you use it by using the dash 2 flag, which means it's supposed to do cdl2 stuff and that currently only works with output. Type cddl, so minus t cddl is, is an option. You need to put there and of course, if you know posix command line argument rules, you can stick this together, so Dash 2, T cddl, is what you will typically use and yeah you.

B

You apply this to one or more input, cddl files and get a combined CDL one file which I have called completed, City delt. Here and often you don't even do that, but you you just pipe the output into another tool, uh so you can generate instances, that's the the cddl GP here or you can verify instances that would be the dash VP or excuse me validators.

B

The right word not verify so this the need for this will go away with respect to the early cddl but, of course, other tools, the the code generation or things like that will still need that. So that's how you use this and um yeah. Let me just to explain the the main components of this and and I will uh start to ask questions to myself a few slides from now, but right now, I'm just presenting what what uh this uh projectile does uh so um semicolon hash include.

B

Does a wholesale inclusion of another CDA file um and the wholesale inclusion is sometimes a good idea, because you, if, when you are developing something uh you want to be aware of unused rules.

B

um So that's why there is an include and not just an import. The input simply gets rid of unused rules and that's exactly what you don't want uh in in a development situation. So if you wrote a rule and then that isn't used, then probably something is not working right.

B

um The import directive operates on an undefined list. So when, when the CDI processor processes a CDA file, it looks at all the rules in there and computes. A set of rule names that are not defined so not defined here means the the current module doesn't Define them and it's also not defined in the Prelude. The Prelude is currently is still the special case. I I hope to be able to explain what the Prelude does.

B

The the thing where, where we're real names like into and float and so on, are defined um so of course, that shouldn't uh turn up in an undefined uh list. So it is somewhat special and that that will probably change at some point um yeah.

B

So uh after processing the the file uh and and removing all uh used names that were defined and all names that are defined in the Prelude, you have a list of names that somebody still has to Define and that's the undefined list and what import does is it looks at the module to be imported and tries to find those names and if, if those names are in a rule in the imported module, then that rule is important, plus everything that is not defined uh in that rule, every additional name that it uses that is not defined, gets added to the undefined list, and we do that until we we have either finished the undefined list or uh there is the it's.

B

There is no resolution for an undefined Christian, a.

A

Brief question: um if the thing that is being imported has internally missing items, will those missing items just leak and create some um some scope that I would consider weird in that it only laterally gets imported by another another input, yeah.

B

A

B

A

All the undefined items from an import need to be resolved.

B

So the the the undefined list of the imported item doesn't leak into the undefined list that will be used by the next import so that there is some isolation here, um but definitely the the way.

B

This isolation, X, is something we we need to discuss in in a few real world examples, but I would expect I'm going to use import on something that already is is somewhat stable and somewhat complete, um so it it knows um how to do this, and what the the one of the next slide says is that the imported module can, of course, itself have Imports again and those are at least conceptually processed before the imported modules. Even looked at so, for instance, RFC 1953, the Cozy algorithms uh defines a few pieces of cddl.

B

That only makes sense if you are using RFC 1952 the Cozy structure standard, and so, if you ever include 1953, you actually want those Rules from 1952., but you don't necessarily want all the other rules of 1952..

B

So the 1953, when you're, only includes those rules from 1952 that are needed to complete the routes that are in 1953.

B

Yeah I I understand the silence, but it really works out quite well in in practice.

B

Okay, um so this is all fine. There is a little additional thing and that's where the namespaces come in, um because the the world that you are importing here may be huge and there may be lots of things defined in there.

B

um It's probably a good idea to be able to name space them and the as clause in the include that can actually be used to say. I, don't want to these rules to be imported as they are, but I want them to get a namespace prefix, which is just something with a DOT ending with a DOT, um and that makes sure that the the rule names uh do not conflict.

B

But you don't have to do that if what you are importing is trivially enough, or you know that they already have taken care of namespacing of using unusual names uh in some way in in there, the module and I'm going to get to this in a moment.

B

Okay- and this is the the slide that actually tries to answer the isolation thing so uh when there, the when a file is referenced by an import or an include um that referenced file is, is import into a process before it's actually going into the the import include process that that initiated all this.

B

um So the the processing of the referenced file happens in isolation. The referencing context does not change the meaning of a referenced module and I think that that's a pretty important um uh property that we want to maintain. So you don't get surprises in what what import actually means.

B

um Based on on uh what the environment, the the import set up, the only thing that is bridging between it towards the undefined list.

B

So this means that indirectly, imported rules are present in the imported module. So when you import 1953, then everything from 1952 that is needed to do 1953 um is uh in the 1953 module as well, and that's not a problem, because we have this. This uh um I'd important redefinition idea in in cddl, so it doesn't hurt if things are defined a few times, but also, if you actually use namespacing, then then you will further increase the the isolation here.

B

um So the the foods actually even have different names, but it means that if you include 1953, um then you actually get to use the Apache info definition that is in 1952. So you don't have to do anything to do that, but because you have to have it to to uh operate on on what 1953 uh defined. But you are not getting all the rest of cozy uh India unless you do a separate import of 1952..

B

I think that that actually addresses Jeremy's concern, uh or there is of Jeremy's observation about the the weirdness of uh doing nested- includes um that that actually might not span a tree but but have a general directed acyclic graph, and so that need that leads to multiple includes and multiple includes, are not either important in C, and they are either important here.

A

So, like diamond, include.

B

Yeah yeah and they are innocuous they also by by name spacing, can be made even more innocuous, so they don't even touch so you can um inflation uh your tree. What's the English word for inflation.

B

Okay so probably.

A

Disentangle I think.

B

Yes, entangle, thank you.

B

um So where do we get those module targets from uh so basically I thought. This is really something that should not be uh nailed down in the module files, because you want to freely move around module files and the modules are not going to know where they are and and where the modules are, that that these modules actually including and if you actually want to do a new version of a module uh for some reason you want to do a bug fix.

B

Then you have to make sure that you don't get the original one, but the new one, and so on um so I thought that the the place where the module targets come from um should be defined in an environment. Variable and that's called cddl include path that works like most the Unix path environment variables. So it's a number of directories that are separated with a colon and Dot, of course, is the current directory and there is a special meaning for the empty string.

B

So the empty string points to the the batteries included space defined by the cdlc module. Now, that's probably something that that we we have to get some agreement on what what's in there, um but as long as we only have one implementation of the CDA process, I think we can ignore this problem and once we get a second, then then we probably want to set up some coordination of Yankee reduction. Something like this, an INR registry um if INR is willing to play along with us.

B

uh That makes sure that these batteries included are compatible between different implementations.

B

So right now that directory includes cddl from all uh rfcs that were done in the XML age. So from 8650 onwards, I haven't actually looked if earlier, uh RCS have interesting, cddl, so I I would love to get feedback on that and, of course doing something like like the 1771 example that is in 8610 um would would also be a nice thing to include there, even if it's not totally official, but at least it has been published in an RC.

B

So most of the things come from your your file system and a few things come from of the batteries included, directory and uh deliberately I didn't try to address uh sources from the web um because that's complicated and that's why you see an empty checkbox beside the to-door called scraping here, because essentially I think we will need to do scraping on on essentially everything that we referenced, including uh internet draft.

B

So uh the the three things that I think we need to have is Ayana uh internet drafts and some form of GitHub import. uh So you actually can point to someone else's project which of course doesn't need to be limited to GitHub it. Certainly it should include other places like gitlab, um but yeah manufacturers might have sites with that as well and so on. So here we are in getting catalog uh space here and probably need to do the same kind of thinking.

B

Yeah, so the Cydia England path really is the most important part here and, and the rest will come over time. So this is the the Portside, and now we maybe should quickly talk about the expert side. So how do you actually export cddl from an internet draft or RFC and the the default export is to to Simply take all source code elements that have the type cddl extract, the text from them and concatenate them?

B

This usually gives you something that is useful. It may give you a little bit more if the RFC hasn't taken care of separating its examples from its normative content.

B

um So that there is something called add name, the name attribute in in the source code element of RFC XML, that you can use to sort source code, uh fragments into different buckets, and we just need at least the convention uh to to for how to use the ad name uh for uh doing this.

B

Sorting or you will have to actually manually, scrape and and so for, for the like 20 RCS that that I have in the batteries included directory um about half of them had some manual uh tampering uh to make sure what we get is actually useful.

B

So often you will actually not just throw out stuff, but you actually want to have separate modules and then, of course, these separate modules need names um and yeah. We should have a convention. The name, of course, should derive from the ad name uh attribute. So so you get some control over that um yeah and and I also added uh some some uh semicolon hash Imports uh to some of the CGA files.

B

I talked about what you need for 1953, so that that is not useful unless you also include 1952., and uh so this is kind of automated here, and we certainly should automate that in any future uh cdda we export. So we don't have to uh do this manually.

B

Okay, so we we, the checkbox, is empty because we need a convention for how to export something from a doctor. And um one question of course is: do we actually need a directive uh to control that um so um I said we? We can use the name attribute to sort something into a different uh bucket, but maybe we want to to annotate this with some additional information. So it's clear how this is imported, so the export statement that that's in the second line of of the slide is yeah that can be a convention.

B

We don't have to write this into the uh document, but there may be other things that are a little bit harder to uh infer from from what's in the document. So maybe we need something like an export. A director fee, I'm I, don't know of any need yet, but we might come up with something so a different thing that that an export directive could do or maybe that's a really two different directors um would be to identify those names in a library module that actually are meant to be imported.

B

So, for instance, well, I didn't write all of them, but in the Cozy module in 1952, of course, Cosi key and cozy encrypt and cause encrypt 0 are really the ones that are meant to be imported and not label or values. This those are not definitions that uh make any sense on its own. They make sense in the context of a cosine module, but they probably need to be namespaced away. So your label and your values are not not impacted by possibly inconsistent definition.

B

So again, I'm not sure we need this. I didn't need this in in the current experiments, uh but it's something we could do if we find out that that this is useful.

B

Okay, so this is scraping.

B

Documents scraping Ayana is harder to do because Canada has no defined API, so whatever scripting we do is probably going to be tentative and we are going to rework this at some time time when Ayana gets an API in London, I talked to some INR people and they seem to to have do you find API on the radar now so the the thought we should be having some discussions in this year and help them Define the requirements uh for that. On the other hand, um an experiment is only a good experiment.

B

If we can do something now, um so I was trying to come up with something that that actually makes sense with the the undefined API that Ayanna currently has, and, for instance, for cozy, which is always my go-to example.

B

There is something called analog assignments called sikosi.xml, and in that you can look for registry elements that have record elements that have value elements, and you have to look at the the ID attribute of the registry to find out how that registry is actually called, and this could be it's extracted using something like a DOT INR operator.

B

um I'm, not phrasing. This as a directive, because that director would have to to interact in a weird way with uh names, but that's also something that we definitely can can change here. uh So the the three variables here are cozy algorithms and value, so uh the the values you find in the Cozy algorithms registry would go into a type Choice here. That is essentially the content of the dot Anna registry. So, of course the algorithm would mean all values that are registered uh in the Cozy algorithms registry.

B

And, of course, the the meaning of the changes over time. So um that's another reason not to do this at the directive.

A

Is this is only about limiting values, or would this also give us some means of annotating, um say a numeric value with with its label or with another field, because these annotations are something that's I think also on the radar for for cddl2 and getting them from their convenient.

B

That's very good comment. Yes,.

B

Yeah, so this is here um essentially focusing on the selection, but they may need to be some transformation, some projection and so on, um and uh we we have to collect, use cases like the one you you just sketched um to to make sure we have good mechanisms here.

B

Okay, um scrapping internet drafts is I, think I'm mostly covered that, and definitely we will only support this for XML submitted IDs. So uh those people who want to write the acidity drafts in Android uh they will have to do their own scrapping, um that's the thing.

B

Yes, yes, I actually worked with someone who was it getting an old um RFC back to Android. uh So this still exists, yes, wow. So it's a lot of work for the Secretariat, because the Secretariat then has to do the transformation from Android to XML uh since RFC 8650. But yes, it does exist and there are also people writing their uh in terms of their world, so they're in Greenwood um yeah.

B

So when we reference draft- and we have the usual issue- whether we actually want to reference a specific revision to get a stable uh outcome or reference, the newest revision to get yeah the newest outcome, so that will need to be specified, probably by simply attaching the dash 10 revision number to the import name. And then you know it's a specific version, except for the one internet draft that actually uh describes how to delete code points, 8, 9 and 10 from bgp, which is called draft, something delete, 8-9-10 and that is yeah.

B

Unfortunately, it doesn't have CDL in it. So we don't have to discuss this.

B

um There are, of course, also internet drafts that have multiple modules in them, so the same considerations about using the name attribute for for source code elements and and defining a good convention for that applies here.

B

Next question scraping other sources. Luckily, the other organizations that actually use cddl um all are stuck in some some weird IPR issues, so we haven't needed to do this, but I'm hearing that this is changing now, um so we we might have some some interesting other seos that expose their cddl and we may want to scrape that.

B

So if you encounter some need to do that, please tell the list or tell me.

B

Now next name spacing so um the the import statement here, import, RFC 1952 as cozy just is an instruction to the CDA processor to actually prefix all names that are in the imported rules that are not preload names with cozy Dutch.

B

So the only exception is Prelude names, Christian. Looking.

A

At what was done in 9252, um the idea back then might have been to do something like um from from RFC 1952 import. Cozy underscore wildcard um question is: does it still make sense to follow that route, or do we just except that this is a road not taken.

B

B

To select from a collection of names is somewhat brittle.

B

It might work out here.

B

But I would be a little bit hesitant to make this a major feature, because this works until it doesn't and you might not have much control over when it stops working I mean for stable rocs. It will stay the same, but if you are talking about internet drafts or other sdos documents uh that you can get your nasty surprises, but yeah I think it's a good idea to to think about wildcarding as a way uh as an alternative way to to um get the the important properties of namespacing without having to prepend this cozy Dot.

B

But we still have to do something about uh the rules that are in the transitive closure, uh like the the label rule that I just talked about so RFC 1952 has a label Rule, and we cannot just dump this into another uh cddl spec, because they are likely going to have a label as well, and that might be subtly differently defined.

B

So we we have to do something here uh and and uh any case, so that's actually something that I also talked about on. uh No, it's three slides ahead, I I, get to that.

A

um Just brief question does does: do we really have an issue with that label room because that's already in that should be covered by the undefined list so late we haven't. A new document, has a live defined label so that gets used and uh if the, if the inputs are as hygienic as you mentioned, then, if it has label in there it would still refer to the label. That is not addressable from the outside um of of the internal important.

B

Yeah, but the isolation is not that great. So if you actually decide to import a rule, you import the rule, and this means that after import, RFC 1952 without an escrow C, you have a label in your root set.

A

Even even if it, even if it's not on the undefined list,.

B

Well, it gets on the undefined list because something in 1952 that you are uh explicitly importing via you and Define this uh uses it.

A

Yeah, but only for the scope.

B

There are no scopes, I mean the the result is a cda1 fire and CDA one file doesn't have Scopes, so any scoping we can do we have to achieve by by using namespacing.

A

A

If I would right now red document uh that says, label label is you in and then okay, that's probably bad, because label is actually label is my my funny stuff and then I import, RFC, 1952.

B

Then you're getting a conflict.

A

B

Because the transitive closure of of your undefined references is what gets dumped into your CDL file. So you get another definition for label with which conflicts with what you have.

A

Wouldn't the wouldn't the cddr processor be completely within its right to re-enable rename this label to underscore underscore I. Don't have a prefix for this, but I still give clear, prefix underscore label and and then have kind of it still can have Scopes. And if the thing it exports to doesn't have Scopes it might need to get creative. But.

B

All I'm saying is that it's probably easier to give the the person who's doing the importing here a way to to decide what that scope name is, and then yourself.

B

So the the first line, uh the the semicolon hash import uh line with the as cozy. That makes sure that you won't have that conflict.

A

Yeah, but that also means that, if I'm using any document that is not already a set in stone as an RFC, I will have to go for as cozy, because a new document, if a new document revision, even if it's like otherwise completely compatible, could always introduce some name. That I also happen to be using, so that that would make not known as Imports.

A

B

Completely hazardous.

A

B

Yes, okay, but you may not care.

B

Yeah so um I think the the the convention should be to use this as the stuff, and um there may be reasons why you don't need that. uh But that's probably because you know exactly what you are importing uh there. So the the coupling gets higher. If you remove the s cause.

B

So there is another aspect about this name spacing uh that uses this dot name character, which we actually put into CDL to be able to do something like this um about seven years ago um and um of course, there are specs out there that have something, like course, with Dove dot something uh so they already are using such a name character and that would might lead to slightly uh and weirdly crossword dot, crossword dot, something in the output CDA, so yeah, um but I.

B

Think in in the end, since the Importer controls the prefix that is being added. The the likelihood of of uh collisions here is uh pretty small, except maybe for innocuous collisions which come from uh importing something in a diamond and that's already handled by The Identical redefinition rule so that that doesn't hurt.

B

Yeah and and I think the important property is that you get told about things. You are doing that don't work.

B

Okay uh racing through this now um the the uh objectives uh for for having and, as is to namespace of the internal names, what we just discussed and the objective of not using, as is to have uh uncut cluttered uh names, and maybe there actually is something in between which is this weird statement here, which says we impart RFC 1952 as cozy, except the name kosiki. That is not namespaced, so that that may be a little bit weird. But um that would solve the problem that you don't want to to get this.

B

These names cluttered and, of course we could have some wild card rule here um that that would be a place where I would have less of a problem with wild cards.

B

Okay, I haven't implemented this from thing uh yet, but it would be pretty easy to do that.

B

Okay, um I have two more slides um one is uh we have to have bike sheds. No protocol is complete without bike, shots and I think I found the first one.

B

uh Should there be a space between the semicolon hash and the word import or include so I'm asking you to take home this style questions uh and and when the opportunity hits give me an answer, but I'm sure there will be a few other things that that I just chose something for which has a much better choice that I'm, just not aware of so please do think about that as well and finally slide.

B

Almost all of what I said today works with ABN f. The only thing that that really is difficult is that ABN F doesn't allow dot in the name. So we will have to come up with some other conventions like three dashes or whatever um yeah, but otherwise we can do all this for ABN F.

B

Obviously, the the processor is a completely new implementation, because the processor is uh cognizant about the the structure of CDA, so the processor for a b and F will have to do different things for a b and F, but I think the the principles actually work and being able to reference a b and F and possibly have that. Namespaced is really really used for that. So almost all documents that do something with ABN F have some references from 5234.

B

uh Like digits and and so on, and we should um probably make that easier, but right now, of course, we are not going to try to actually change IBM ABN F, but just to do this for the ABN F support in in cgda. So this would then at some point actually update RFC 9165..

B

You also have some other ABN F modulations in in on a GitHub repo that I think I will be telling people about soon foreign.

A

Questions comments.

D

Yeah you had one small thing uh and yeah I need to look into this a bit more to understand it better. To give some some more, but one thing I had in mind was the these as keyboard that could be used for both to import and include statements. Is that correcting.

B

Can you say it again, the.

D

Answer the ass term or keyboard could that be combined with both include and import? Yes,.

B

Yes, originally I only wanted to provide it for import, but then I found it's easier to implement it for both.

D

Okay, but that that then okay, fair enough, then I get this so it's usable for both okay yeah. Thank.

C

You my comment: is it's a lot to digest.

D

B

It's incredibly simple, I was really surprised, but it's like 80 lines of code or something like that and.

D

It's nice, though, yeah.

C

For me, I didn't mean for the processor.

A

um I I also have a comment um on the on the topic of import as um I. Would one hope that I have for cddl2 is that it will help me match the match and match an item that has cddl?

A

That is that I know to follow some cdtl with code that I have um and when items get when cdtl gets names, get changed through that import I may want to still have access to that original name, because um it that that name might be somewhere in the code because I'm, for example, parsing it into some cozy structure. That knows all its items by the original name.

A

um I I'd assume that if my processor, where cddl2 aware it wouldn't need to go through the expansion and would thus see the items somehow in their original form, but I'm not sure how that would work. The way I'm thinking is like, like you like in in I, don't know if it works in C. Definitely it works in Rust, where you can have a function, but also tell it that its name for the purposes of linking is this and that roughly that that mindset I, don't.

B

Out of that is covered by the fact that something like import from uh will simply generate Foo equals cosine.fu rules because that's much simpler than uh trying to to change the imported module to use the unnamed spaced name everywhere.

B

um So the information will be in the rules, uh but yeah I think it would be nice to be able to recognize automatically added namespace prefixes. So maybe we need a different convention like dot dot and not just dot.

A

Dot dot could still be present, but maybe maybe the programmatic approach is like already good if it's, if it's somewhere in the metadata, I'm good.

B

You would warn people that it's not the great idea, like not the great idea to start very real names with underscore and see.

A

Like like everyone, he did that.

B

Yeah but I mean everybody who works with C knows that they will get be get burnt so.

A

Yeah right, you would hope so there.

B

May still be a good reason to want to get the entrepreneur. Yes,.

B

Yeah, that's an interesting point. I didn't think much about uh yet. So thank you for that.

A

I guess we'll come back to that when the topic of of back annotating siboro comes into well. Annotating cable comes up.

A

Okay, any more questions on on this anything for any other business.

A

uh Here again, there's one very brief: sorry um customize that yeah.

B

I just wanted to say that my uh objective here was to get you to actually play with the thing and install cdlc in a recent version and start using it in your project. So I get some more feedback about how well it works, it works and where it actually starts hurting.

A

Speaking of how well it works um when I don't say, cddl I get some EST output. Is that stable and well documented and or has a plan forward, because that looks really nice and I may want to I'd, be much rather pass this than cdtl exactly.

B

That's one of the reasons: I wrote the tools because people want Json and and no don't want to write pauses. um So it's documented. The freezer document actually has the section about that.

B

um It's also actually an AST and I found that I actually want to process uh rules before actually working on them. So there is a second form that you get with a dash r uh flag for Syria at Sea and that's actually much more useful, but also it's not documented. So what's that funny?

B

What's that t flag, Dash, r, cdlc-r.

B

And then you get another form of Json that, for instance, has parameters of generics uh resolved. So you you no longer get the name of the perimeter which, which you really don't care about, because it might be conflicting with some other names, but you get a number that is local to the rule. That is using the generic, and these are the little things that are done with the dash R and it's the dash R form that uh I actually know how to uh do these.

B

These cdl2 things on, and that is also supported by the CDL writer in in the tool.

A

What language is that? Because it has a equals greater than, and it doesn't look like Json, but it doesn't look like that. Sibo types take notation either.

B

B

Probably see it in Json, oh yep, thanks.

B

Yeah one action point for me: write this up: it's, of course not it's even simpler than than the AST stuff, but I have to write it down.

A

Great, thank you. Everyone I think. That's it for today see you all in two weeks well, actually in one week, because that's Seaboard but yeah have a nice evening. Yeah.

D

C

You thank you bye-bye.