From: Russ Allbery (rra@stanford.edu)
Date: Tue Apr 24 2001 - 05:35:51 CDT
Florian Weimer <fw@deneb.enyo.de> writes:
> Russ Allbery <rra@stanford.edu> writes:
>> I think the best thing we can say in our draft is that the header must
>> either be absent or contain accurate protocol-specific information
>> appropriate for the means by which the reader retrieved the article.
> Perhaps it makes sense to provide a namespace to be used such protocols
> ('NNTP-*' doesn't work because of existing headers, I think, maybe
> 'Transmission-*' is better), so that these headers can be recognized by
> relaying and serving agents when they are not supported by them.
> A different option would be a 'Transmission-Headers: start' which
> indicates that the following headers are not related to the article
> itself, but to the transmission protocol being used.
I think this is another solution based on the idea that Xref generalizes,
sort of like the current local headers stuff. My counter-argument is that
I think Xref is a unique problem, and would like to see an example of some
other header with similar properties before we go to the work of designing
someting into the standard to deal with it.
Xref is a major wart, yes, but generalizing it doesn't make it any less of
a wart and if anything just encourages people to develop more warts.
There shouldn't be transport-specific headers in news articles in the
first place if it can possibly be avoided; Xref could have lived just fine
in overview had XREPLIC or something like it won over Xref slaving as a
method of replication.
-- Russ Allbery (rra@stanford.edu) <http://www.eyrie.org/~eagle/>