jkh@meepmeep.pcs.com (Jordan K. Hubbard) (08/28/89)
I just converted over to NNTP after using nnspool.el and an NFS mount (lose lose) for /usr/spool. Now that I'm using nntp, response is much snappier, but I'm left wondering several things: (from nntpd) XHDR Retrieve a single header line from a range of articles. 1. Why doesn't GNUS take advantage of the XHDR feature of nntpd 1.5 to retrieve article headers? Looking at the code now, it appears that some really grotty stuff is done to avoid choking up nntp after ramming "n" HEAD requests down its throat. All this would be unnecessary if you used XHDR, if supported. 2. Given that most users spend a lot more time reading articles than fetching them, it seems that GNUS could be a lot smarter about pre-fetching the next one while you're reading. This would speed up interactive response a lot, obviously. Likewise, you could pre-fetch the headers for the next group while reading the last article (or last couple of articles) in a group. A proper implementation of this could result in a perceived response time of zero! I'm actually thinking of hacking the article pre-fetch directly into the nntpd, so as to make it available to other programs, like xrn. The "get next group's headers" wouldn't be practical there, however, as nntpd has no way of knowing the next group you've subscribed to. This would be a great potential improvement for GNUS. So, how 'bout it? I have GNUS version 3.11. Is this recent enough to consider hacking these changes in myself, if noone else will? Jordan PCS Computer Systeme GmbH, Munich, West Germany UUCP: pyramid!pcsbst!jkh jkh@meepmeep.pcs.com EUNET: unido!pcsbst!jkh ARPA: jkh@violet.berkeley.edu or hubbard@decwrl.dec.com