Hi Tony, I’m personally fine with it, and since we did not receive any negative comments yet, the Thrift community is most likely fine as well.
Only thing to keep in mind is that this is all FOSS here so everything is voluntary work. What I can tell is that the basic protocols (i.e. binary, compact, json) and the basic extras like TFramedTransport and some others are stable and productively used for years. Other stuff like THeader for example is relatively new and/or not widely adopted yet even in the Thrift libraries, but the wire formats are unlikely to change during that process once the first reference implementation(s) are completed. Also I would like to point out the written docs we have: a.. http://thrift.apache.org/static/files/thrift-20070401.pdf b.. https://www.manning.com/books/programmers-guide-to-apache-thrift c.. https://issues.apache.org/jira/browse/THRIFT-1550 Have fun, JensG From: Antoni Przygienda Sent: Thursday, November 11, 2021 6:20 PM To: Jens Geyer ; Thrift-Dev Subject: Re: Stable thrift reference for RIFT IETF standard ... Hey Jens, thanks for your answer, yes, it kind of does point me in the right direction. RFC3967 basically says (that was the doc, you probably clicked on another one) that IETF needs pretty permissive license before publishing anything. Given how permissive 2.0 is it should work but it’s up to a liason/lawyers to hammer out. So, from the technica, head-of-thrift perspective do I read correctly that you feel that we could work on a stable information RFC to standardize an “official” thrift spec that other protocols/services could use since you seem to have a pretty loose set of documents/records only that will be hard to reference as the “thrift protocol”? As example QUIC has been standardized that way out of Chrome basically to its benefit AFAIS 😉 https://datatracker.ietf.org/doc/html/rfc9000 As to gathering documents, pouring them into a proper informational rfc standard I’m more than happy to do that, get folks to work on it so now work involved from the thrift community side except giving us pointers to all the info that will affect the rfc. And of course reading/commenting on the work If you’re ok with that then I’ll probably try to talk to ASF heads/legal & IETF to start a liason as in “we’re fine thrift being gathered for IETF purposes into an informational RFC standard” a.. Tony From: Jens Geyer <je...@apache.org> Reply to: Jens Geyer <je...@apache.org> Date: Wednesday, 10 November 2021 at 22:46 To: Thrift-Dev <dev@thrift.apache.org> Subject: Fw: Stable thrift reference for RIFT IETF standard ... [External Email. Be cautious of content] Hi, Yes, I have seen that mail as well: https://mail-archives.apache.org/mod_mbox/thrift-user/202107.mbox/%3c03439030-2f4c-4a2a-ae21-5daeaac2d...@contoso.com%3E The only spec we have is what is in the source tree, you found it already. https://github.com/apache/thrift/tree/master/doc/specs Additionally there may be some things buried in the various JIRA tickets that’'s not yet in the github repo. Plus we have some README like stuff in most of the lib folders, covering language specific stuff. > Pls Read section 10 carefully since we’d need to get a license from apache > under the precise terms defined there. To be honest, I have no idea what you are talking about here. Especially because there is no section 10 in that linked document. All of the ASFs open source is under (big surprise) Apache 2.0 license, which is pretty permissive. If it is legal advice you’re after I would point you to the ASFs legal mailing list instead. Does that answer your question(s)? Have fun, JensG From: Antoni Przygienda Sent: Monday, November 8, 2021 5:12 PM To: je...@apache.org Subject: Stable thrift reference for RIFT IETF standard ... Hey Jens, writing you as head of Thrift in Apache as far I figured and not having had much success on the user’s list … So, IETF is in last steps to publish a pretty big specification of a modern routing protocol based on thrift as IDL (consciously driven by me, IME thrift is head-above-shoulder as IDL over other, more “modern” stuff and the compiler/libraries quality is bare to none). https://www.ietf.org/archive/id/draft-atlas-external-normref-00.txt Protocol has @ least 2 implementations by now & going into deployment. To publish an IETF (Internet Engineering Task Force, the guys that standardized the “Internet” and “IP” 😉 standard we need a “stable reference” which is roughly what https://www.ietf.org/archive/id/draft-atlas-external-normref-00.txt describes. RFC3967 pretty much forces the republication of an open standard referred to by an RFC as as independent stream RFC or otherwise a very “stable” reference. Pls Read section 10 carefully since we’d need to get a license from apache under the precise terms defined there. Now, thrift is bits loose in this respect (if we’re talking tight world-wide stable standards reference) with the only stable doc I found on ..org/docs/idl While on the github there is bunch more documents describing e.g. header format etc which are kind of integral part of the spec. Question is, would thrift benefit from being poured into an official RFC in IETF (informational track referenced then first by RIFT specification). IETF is more than happy to support and I’m more than willing to put in the work (or find folks doing it). What would be the relation to apache foundation (do we need a liason or something?). If not, how could we stabilize the spec to a tight, one document version with a stable publish point? Looking fwd’ to your answer … a.. Tony Juniper Business Use Only Juniper Business Use Only
