Hi Paul,
Responses in-line.
On Wed, Jan 24, 2024 at 6:44 PM Paul Hoffman <phoffman@xxxxxxxxxx> wrote:
On 24 Jan 2024, at 8:30, Ted Hardie wrote:
> Seeing Eliot and Russ's points, I think that this could be the basis of a
> valuable change, but that it may need additional work. We have changed our
> system to recognize that these drafts no longer disappear, because it was
> clear that external archives had made that a fait accompli. But this is
> asking for something different, declaring that the TTL field has no meaning
> at all.
This is kind of a mis-statement about what the draft says. The draft says that there is no "TTL field", not that it exists but has no meaning.
You're right. I should have phrased this differently.
> I think we can probably do better, by making it more variable, allowing any
> party to set it shorter than the default 6 months, but requiring some
> action beyond the author's preference to set it longer than a specific
> amount. With a specific set of circumstances, I can easily see one of the
> values being "no expiry". If there are concerns that this does not fit all
> cases, the answer may be to allow both that and other choices, using a set
> of processes which need to be ironed out.
It sounds like you are proposing keeping the concept of expiration, but making it have variable time. Maybe I'm being uncreative, but I don't see how this could possibly be implemented given that the expiration time is instantiated in the draft itself.
Yes, I think one alternative to this that Eliot and Russ's points suggest is that this be variable. I would suggest that this continue to be marked in the draft. To sketch this out a little further:
The default expiry of 6 months stays the same.
Anyone can for any reason declare an expiry of shorter than 6 months.
Any expiry beyond 6 months must be approved by the relevant part of the IETF/IAB/IRTF. In the case of standards track IETF drafts, that would be delegated to the working group chairs for the working group. There would need to be a designated stuckee for other types of drafts.
For what it is worth, I agree with you and Martin that there are cases where the expiry should be effectively absent (so an infinite time to live). For drafts which are used to request IANA code points, for example, expiration within six months doesn't make a lot of sense. But there are drafts for which it does still make sense, and rather than trying to prioritize one over the other, I suggest we do both (insert meme of girl being lifted on the shoulders, if you like).
Can you model out how you thought the variable expiration time would be known to the reader of the draft, or in the many places that the draft is stored?
I hope the sketch above answers this question.
regards,
Ted
--Paul Hoffman
