Registries Breakout -- 18 Sep 2019

<dsinger> https://www.w3.org/2019/Talks/TPAC/ac-registries/Overview.html#start

<scribe> Scribe: Travis

(dsinger presents slides on Registries) link?

<tantek> hmm I wouldn't expect this to fly because the reason we have so many ways of doing registries is to get rid of all this process / bureaucracy

<tantek> TBH this is even more onerous than IANA policies

<florian> URL to slides: https://www.w3.org/2019/Talks/TPAC/ac-registries/

dsinger: so which things can you publish together.
... registry definitions, referencing documents, rules, etc...

"W3C Registry" as a new kind of publication

<tantek> oh dear

<fantasai> tantek, not sure why you think it's more onerous than IANA; you could literally set up a form on a server somewhere that accepts submissions and publishes them directly into the registry on /TR via Echidna automatically

<fantasai> you probably want to do some spam-checking, but otherwise that's the minimum requirement

dsinger: Finally ready to hear from the room!

<tantek> "you could" but no one has. I'd rather see examples of success than a hypothetical untested process proposal for registries

nigel: definition of registries looked like data... but is it a document?
... does the intro say "this is a registry" etc.

<tantek> what's CCG?

dsinger: illustrates an example merging the rules + data
... updates can occur if you obey the rules in the rule section

<manu> https://w3c-ccg.github.io/did-method-registry/#the-registry

<fantasai> dsinger projects https://www.w3.org/TR/timing-entrytypes-registry/

nigel: Is this a new state between informative and normative?

dsinger: Can't we just publish as a REC? Others say, no a new thing... don't care either way.
... want to ensure whatever we do has had community review.

<manu> https://w3c-ccg.github.io/vc-extension-registry/#the-registry

dsinger: I think REC track does that well. Exclusion opp. is empty, but that's harmless.

florian: to publish all parts separately... then...
... various parts need different process

<Zakim> nigel, you wanted to ask if informative text is allowed - isn't a registry simply a document?

florian: if we require them all in a REC, then we have a REC process already!

<tantek> dsinger we need queue management of q+ vs mics

florian: I prefer balance.
... Rules + Registry together AND both inside a sepc.
... Don't think it makes sense so publish the rules from the data.

dsinger: Well, the largest registry does it that way...

florian: it's subjective.
... if allow it in the spec, need a rule to allow content of the table to be updated in place.
... many of the rules are there to suppport patent rules...
... we can then adapt our process to match

yoshiro: IANA is biggest registry... why not use IANA for the registry?
... they have proceedure already defined. Process could be applied in W3C too. Reuse is good practice.

florian: W3C does not have to manage all registries in the world. Often W3C does normative references...
... for people that want to do a registry at W3C we don't have a good answer.

dsinger: We *could* ride along on IANA and have them help us...
... but seems like a formalization problem (not that we don't know how to do it)

joshiro: can be a discussion between SDOs.

<Zakim> manu, you wanted to talk about CCG registries, who gets to update, how are things maintained?

manu: CCG currently has a set of informal registires
... extensions to VC spec. WG had to figure out who would manage
... Decided that CCG (since it would continue to exist) would manage the registry
... same in several other examples.
... so CCG now has a bunch of registries.
... But process I see looks quite heavyweight
... Turns out we have 32 new DIDs
... every week!
... would be quite a burden!

dsinger: To add a new row is as lightweight as you define.
... We won't put requirements on the rules... we just want you to follow the rules that you had approved.

fantasai: You can assign a custodian... you could link it to a form submit button, etc.

dsinger: As long as it's fully acceptable to the community.
... The WG can setup the level of review that is necessary or appropriate.

manu: The CCG made a process to put the rule definition inline in the document.
... Rules and values are together in one document.

tantek: Also got an impression that the process would be heavy!
... have seen many registries go to other places because they say a heavy process and decided not to go there (so created their own registry)
... happy with what I just heard.

dsinger: So far, havent come up with anything we think needs to be required in the rules for a registry.
... If you Github or something else, you just need to make sure the published values go into TR/ (W3C)

tantek: That may need more explanation... pushing something to TR/ seems like a big barrier.

fantasai: we have an auto publishing thing (Echidna)? This makes it easier..

<tantek> Echidna is still VERY confusing

<tantek> and the setup is definitely not easy

fantasai: you could set that up to work for your registry.

tantek: Please provide the example for how to do this!!! Don't make it someone else's job to discover it.

dsinger: Have asked, what does publishing to TR/ get you?

tantek: I'm not asking to see the benefits...

dsinger: Let me try...
... Formal URL, not Github
... Backed up (redundant) w/history
... Copies only happen if you obeyed the rules...
... if the copy is automatic, then...?

tantek: Github is just fine for folks outside of this room (vs. w3.org)

tobias: thoughts... should verifiable credentials be split out? what do you think

dsinger: Have a large registry that is broken up in multiple tables, but it is a single registry... can be as complex as possible.

Robert: Exclusions... simplest I can imagine is a 3-value registry.
... what if they change one of the values to mean something totally different?
... if the registry value changes, what happens

dsinger: A rule should be created--that you can or probably CANNOT change the meaning of existing values.

Robert: is there a mechanism to redirect values of a registry into another document...

dsinger: As long as its defined as a registry, multiple can refer to it...

fantasai: Example: 5 specs want to have a dir prop that has these values in the registry... you reference it.
... If you want to auto-import the values into the document, you write a script for that.

Robert: Transclusion...

fantasai: Can't change documents in-place... if you copy from another document, you'll need to make a copy of your own document....

Robert: you just want to link to the registry values, but not have to scroll down to teh value section...

fantasai: we have <a href=...> :-)

dsinger: If it's in the spec, then the spec will get frequent changes, but they only come from updates to the table.

florian: If it's used by multiple specs, then just have (missed it)

dsinger: Make the rules the least-restrictive as possible, and then see what happens (that's my style)
... others prefer more structure.
... If you're going to have 3 values, just put in the document!

nigel: You should be careful about those values changing.. maybe that shouldn't be a registry

dsinger: if its a small set of values...

nigel: What if the WG might stop existing... and there are updates?

dsinger: Might be cautious to add a rule that says the WG should do the update...

florian: Can indicate who should do the update (can be CG, somewhere else..)

nigel: So, you'll need to get permission for whomever will maintain the registry.
... that should probably be a rule.
... if it says "the team" might update the registry, and it gets a lot of new values, then don't annoy the team. (get permission first)

Drummond: Hi folks!
... I'm in DID
... A 'registry' in DID is a database
... it's a read/write thing
... I'm sharing a perspective here.
... Just want to let you know this was pretty confusing when I walked in!
... Devs will want to write software against the values in the registry.

florian: We don't need to have registries that can only be read by humans. Could be made machine-readable...
... Examples: you might publish both human and machine readable version

<tantek> My other database is a <table>

florian: or one or the other.

dsinger: The spec could point to the Github repo... the machine version can point to the other location.
... want to ensure this fits within the community's needs.

Drummond: There could be more dynamic registries.. could be process overload!

florian: Not sure if we need to run software services to handle things. It's just a file we need to publish.
... We can maintain a file on a server. That's cool.
... Don't put it in MySQL Lite 2.3.a

fantasai: No reason not to [also] publish your registry elsewhere in a more usable form

dsinger: Am familiar with systems that only provide a search/lookup system. That's probably not going to work.

<tantek> browsers can display the HTML spec, browser can display your registry as an HTML <table>

florian: It really must exist in some form of linear format to be OK.

<Zakim> nigel, you wanted to ask who

drummond: My point, many things that are registres are usually database (with programmatic access).
... would love to see W3C move toward programmable web.
... it's a dream!

<kaz> EmotionML vocabulary Note

Kazuyuki: Example of a registry above
... group was closed, and the registry not maintained.

<azaroth_> Another example from a couple years ago: https://www.w3.org/TR/annotation-model/#motivation-and-purpose

Kazuyuki: we need to think about who will maintain it when it's setup.

florian: Two downsides with W3C Note.

<nigel> TTWG has also used WG Notes for registries

florian: You write rules for the update

<nigel> TTWG Update policy is that WG Consensus is required

florian: Don't want to be able to update the rules in a doc update!

JoeAndrieu: This is GREAT!!! Where were you before!!!???
... I second Tantek's concern about publishing to a TR... bumped into process issues.
... Now that I've done it, I know, but please make it better for the future!

fantasai: Yeah... first time's the hardest.

florian: There are multiple steps.. there's the process AND the tooling.
... need documentation, etc.
... focus here is on governance

<kaz> s|it's set up|it's setup (as Nigel also mentioned) in addition to the style/mechanism of the possible registry|

JoeAndrieu: wiki mentions a lightweight group that's not a CG or WG...
... would love to see this explored.
... don't see our group have the moral authority to handle the registries

dsinger: need guideance on what to say about how the registries get changed. A review expert? Review group?

JoeAndrieu: have already changed the registry.. not just data, but also rule sets
... don't want to do that in CCG.

florian: Sometime the registry is just used to avoid collisions.
... in which case, anyone could do it! Very little process is needed.
... might be interesting to formalize the patterns.

dsinger: also want to avoid duplicate meanings.
... requires checking by a human
... might need to be an expert in the team

JoeAndrieu: had a problem with name-grabs and had to deal with that.

florian: Just want to make sure that changing the rules gets as much review as when they were established

dsinger: Are we done? I've learned a lot.
... next steps:
... create formal process text and get your feedback.
... these folks who've done "proto-registries" can help provide good feedback.

JoeAndrieu: Let's work together to do the transition to grandfather existing registries?

iclelland: On structural changes...
... if I want to delete a column
... what would the rules be like to do this? Esp. if it might be referenced?

dsinger: sounds like: I want to change the registry rules. Will need to delete community approval; chance for review

nigel: Should be good practice never to delete anything from a registry; just mark it as deleted.
... prevents re-capturing the name for something else.

dsinger: That's good practice!

<tantek> tombstoning ☠️

JoeAndrieu: Pattern that might not be supported: in progress draft but using a registry?

dsinger: At some point you need to finalize the registry.
... I want to write this in the process, then to the CG process to allow them to do this...

- DRAFT -

Registries Breakout

18 Sep 2019

Attendees

Contents

Summary of Action Items

Summary of Resolutions

Scribe.perl diagnostic output