Hugo Embrechts wrote: > What I think the problem is. > > The results of xovers are (normally) cached in disk files. When > processing an xover from a newsclient > > - those cached xover-lines which are within the current "active range" > of articles are fetched > - holes in the obtained list are requested from the remote news server > and the returned lines are added to the xover file > - the fetched cached lines and the lines from the news server are > returned to the newsclient > > Therefore expiring of the xover files does not matter. Correct. > I think the missing articles are canceled articles. When the xover of an > article is cached and the article is canceled afterwards, an xover line > for this article will be sent until it falls out of the active range > of articles. If it were only cancels, I doubt anyone would care :) As I said months ago, the major problem is the use of Expire headers which artificially lowers the low water mark to a level where there are big holes in the numbering. > This is inherent to caching. But is indeed ugly. A possible solution > would be to do a LISTGROUP from the remote server before each xover > and ignoring cached xovers that correspond to missing > articles. Evidently there should not be any caching of the results > from the LISTGROUP command. The current caching of listgroup works just fine. What needs to be implemented is the deletion of xover and article files with the benefit of the output of listgroup. > I think this solution would not waste too much bandwidth. I agree, especially since tin uses listgroup for group :) -- Debian GNU/Linux 1.3 is out! ( http://www.debian.org/ ) Email: Herbert Xu ~{PmV>HI~} <herbert@gondor.apana.org.au> Home Page: http://gondor.apana.org.au/~herbert/ PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
