From: Charles Lindsey (chl@clw.cs.man.ac.uk)
Date: Mon Nov 05 2001 - 05:15:12 CST
In <200111021538.fA2FcCjG027283@lima.epix.net> "Forrest J. Cavalier III" <mibsoft@epix.net> writes:
>> So do we agree that feature #2 is worthwhile?
>No. I do not agree.
Actually, I think you are addressing feature #3, not feature #2.
>Benefit:
> For individual readers who wish to revise a post, the need
> will arise so infrequently that few will remember the
> feature is available from the newsreader, nor how to
> use it. So even if the news reader is modern enough
> to support it, it is unlikely to be used.
True. This proposal must stand or fall on whether it does a useful job for
FAQs.
>Cost is mostly in the potential for abuse, as well as
>implementation complexity for server and readers.
Yes, but most of the potential abuses can be committed without this
feature. A malicious person can already post a bogus parody of an existing
FAQ, and all the true author can do about it is to cancel the bogus one
(assuming it was forged in his name, or superseded his genuine one) and
repost the correct one. I don't think this particular abuse is widespread
at the moment, but it is certainly possible.
The question to be asked now is whether introduction of Replaces would
make that abuse easier, or more likely. I think not. I don't think it
actually makes much difference.
>Abuse:
> There are a few well-published message ID's, which appear in
> URLs. Being able to hijack those message ID's after the
> fact is an attractive target for vandals.
OK. Maybe that does make things slightly worse, insofar as the genuine FAQ
writer has to be quick to get his correction in (but the necessity for the
correction is there with or without Replaces).
>I think implementation cost is an issue also.
> Introducing references, pointers, and indirection into any
> index file (like history) is a major complication in implementation,
> clean up tools, audit tools, et al. Ensuring referential
> integrity becomes a major issue. (When does the old
> message-ID expire? e.g.)
Actually, that part of the implementation is not difficult. Expiry happens
as normal.
> The General Assumptions you posted are too restrictive.
> There are other implementations of index files possible.
> (Being able to have duplicate index keys in an index is
> complicated. Overwriting is not always possible when
> records are not equal lengths. And records will not
> be equal lengths for the simple reason that the message
> ID's lengths are not equal.)
Duplicate keys are no problem (actually, there aren't any - you just
arrange that when someone tries to add a key that is already there, you
overwrite it with the new one; the onld entry remains in the History file,
but you never see it unless you scan the whole file, and only Expire and
NEWNEWS ever do that).
Overwriting is only possible if you do certain Cunning Hacks that Russ is
proposing to put in INN, and which lead to fixed length records.
>There are other things that bother me about the feature
>and draft.
> - The draft is silent on what should happen if the Newsgroups
> header is different.
No different from the present Supersedes. Have you a proposal to say
something about that (serious question)?
> - How are newsreaders supposed to draw attention to the
> fact that the message was replaced? Pop up dialogs?
Perhaps. Or some highlighted text at the head of the article. I am sure
Bill Gates will invent something suitably flashy :-( .
>And finally, my standard complaint: this is a completely untested
>feature. It is not documentation of existing practice. Are people
>afraid these ideas will remain unpopular if spread by experimental
>RFC? Is that why they have to be in the draft?
There is an experimental implementation. I agree that this is possible to
be done by a later extension which could be brought in gradually (as
opposed to 'mvgroup' which really had to be got in soon, if it was going
to happen at all). The question is really whether we could propose
something better if we left it alone for now. If not, it may as well stay
in.
-- Charles H. Lindsey ---------At Home, doing my own thing------------------------ Tel: +44 161 436 6131 Fax: +44 161 436 6133 Web: http://www.cs.man.ac.uk/~chl Email: chl@clw.cs.man.ac.uk Snail: 5 Clerewood Ave, CHEADLE, SK8 3JU, U.K. PGP: 2C15F1A9 Fingerprint: 73 6D C2 51 93 A0 01 E7 65 E8 64 7E 14 A4 AB A5