dave@garfield.UUCP (David Janes) (03/23/85)
There has been some discussion recently about how unnecessary
or misused header fields are taking up too much disk space and
costing too much to transmit. However, that is *nothing* compared
to space wasted by 'included' articles. What is needed is some
fast method of finding an article file from a Message-ID in
one's favorite news interface (i.e. rn.) and using that to
get 'included' articles.
A Minor Proposal:
Rnews on seeing the of a Message-ID of a new article, should
put in a dbm file some reference to the file holding the article
(fine, it already does this - it is not the fastest - but it will
probably do.)
Postnews should, on request, include the text of the original article,
suitably indented in an easily identified (ha ha) manner (i.e. '> ' in front
of every line.) The user should delete the lines he doesn't want to
include. Then, when the user posts the article, postnews makes a special
pass through the file, and changes all the included text to lines like
'##<message_id@of-original.article>:startline:numberoflines'
without the quotes around it. For example
'##<2608@garfield.UUCP>:1:5'
will be the directive to include 5 lines starting at line 1 of this article's
*text* in your message.
Rn (or whatever) should then offer two choices -
1. automaticly include the text in the article as it is displayed,
indented to the proper level, etc. It should only allow so many
levels of depth to stop loops and the running out of file descriptors.
2. on the user's request, show the included text, done in a manner
not unlike how rn now handles digests.
In both cases, it should just ignore (i.e. pass on) lines that
don't perfectly match this format.
Comments, flames, counter suggestions?
dave
--
David Janes
The UUCP: {utcsri,ihnp4,allegra,mcvax}!garfield!dave
Mercenary INTERNET: dave@garfield.uucp
Programmer CDNNET: dave@garfield.mun.cdnthomas@utah-gr.UUCP (Spencer W. Thomas) (03/26/85)
In article <2608@garfield.UUCP> dave@garfield.UUCP (David Janes) writes: >What is needed is some >fast method of finding an article file from a Message-ID in >one's favorite news interface (i.e. rn.) and using that to >get 'included' articles. I see two problems with this approach: 1. the referenced message may have expired by the time someone tries to read the message that references it. The slow propagation time of the network ensures that no scheme will be able to prevent this. 2. The referenced message may not have reached the place where someone is trying to read the message the references it. One might call this the "Orphaned Response" problem (and the name also is an existence proof!) The nature of netnews is that no scheme that relies on cross-referencing of messages will work reliably. -- =Spencer ({ihnp4,decvax}!utah-cs!thomas, thomas@utah-cs.ARPA) "The meaning and design of a problem seem not to lie in the solution, but in our working at it incessantly." -- Jung