[rfc-i] Updating one paragraph of RFC 2026 to reflect current practice

Andrew Sullivan ajs at anvilwalrusden.com
Fri May 29 14:40:59 PDT 2015


On Fri, May 29, 2015 at 04:55:20PM -0400, Scott O. Bradner wrote:
> seems to me that referring to a specific version is quite important - the feature that the document assumes
> is in the ID (and is referring the reader to) may get removed in later versions
> an history aside - the language in 2026 comes from the deal done to start the internet draft concept - it would be OK only
> if there IDs basically were never considered “real” documents 
> that was a long  time ago and well overtaken by events

So, here is the difficulty I have with all of this.

At the moment, we have one document series.  It's an archival series,
and when the content of a document that has a particular function
changes, we give it a new identifier.  This feature is what makes it
an archival series.  You identify the particular version of the
document you want with its archival-series identifier.  (This is,
interestingly, a reason why one should not actually use the BCP or STD
numbers in the references: they're not guaranteed stable over time.)

If we make a change such that we acknowledge the actual permanence of
I-Ds and start identifying particular versions of them, then in effect
we are reproducing the archival series feature of the RFC series.  It
seems to me then that this will encourage the already-existing
lamentable practice of referring to I-Ds as "draft RFCs" and also will
tend to encourage ossification of protocol designs even earlier in the
practice.  That is, if you can refer to a particular version of an I-D
and say you implemented that one, one will then have much stronger
arguments during protocol development that there's all this deployed
code out there so the protocol can't change.  If we can make such
specific references, why can't RFPs?  And so on.

The upshot of all of this, AFAICT, is that it turns the I-Ds (which
are now officially an archival document series) into an alternative
publication mechanism for RFCs -- particularly Independent Submission
stream RFCs, which get little more IETF vetting than many I-Ds.

Now, this might all be entailed by the fact that we started permitting
references to I-Ds.  But if we're going to make this change, I think
it is something that the whole community ought to discuss.  It seems
like it has big implications.

Best regards,

Andrew (in case there's any doubt, speaking only as an individual)

Andrew Sullivan
ajs at anvilwalrusden.com

More information about the rfc-interest mailing list