lynx-dev
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: LYNX-DEV News:*


From: WWW server manager
Subject: Re: LYNX-DEV News:*
Date: Sat, 16 Aug 1997 15:03:35 +0100 (BST)

Foteos Macrides wrote:
> Klaus Weide <address@hidden> wrote:
> >On Thu, 14 Aug 1997, Foteos Macrides wrote:
> > [snip]
> >>    The default path for news URLs is '*', which means a linked
> >> list of newsgroups available from your nntp server, obtained by a
> >> LIST NEWSGROUPS nntp command to it.  If your list does not correspond
> >> to your nntp server's news groups, that's surprising, because it's
> >> what your server reported to Lynx.
> >> [snip]
> >
> >For most if not all nntp servers, the reply to a LIST NEWSGROUPS command
> >ist just the contents of a file 'newsgroups' on the server.  That file can
> >be totally out of sync with the newsgroups that are really available at
> >that server.  For example it could be the default list of thousands of
> >worldwide newsgroups when the server carries only a few.
> > [snip]
>
>       If the LIST or LIST ACTIVE command were used, then in parsing
> that reply an XGTITLE request could be issued for each group, but you
> can't count on Xfoo commands being supported by all nntp servers (that
> one probably is by now), and it would be a lot more overhead.  It would
> be better to pester the sysadmin to arrange for the LIST NEWSGROUPS to
> return valid information, rather than accepting that some hoaky procedure
> for reading a file that isn't maintained such that it's contents contain
> what would be a valid reply is OK. :)

Several comments, speaking as both a news and web admin, and lynx installer...

(1) The "newsgroups" file returned by "LIST NEWSGROUPS" (from those servers
which support it; it's an ad hoc NNTP extension, not as yet standardised) is
purely a list of groups descriptions for those groups which have them. De
facto it is not and most likely never will be a complete list of groups
available on the server, since many alt groups (in particular) do not have 
descriptions.

(2) The definitive list of groups is what you get from LIST (or LIST ACTIVE, 
a non-standard equivalent).

(3) While implementation dependent, I suspect that sending XGTITLE for
each group in turn would typically result in a linear search - expected to 
be an infrequent operation, not something a client is expected to do in bulk 
for a list of 20-30000 groups. Slow for the user and very inefficient for
the server. In general, it's probably best to choose a bulk retrieval of a 
moderate amount of data over a large number of requests for trivial amounts 
of information.

The "right" approach for getting details of available groups and their 
descriptions (assuming you need/use the descriptions, I've not looked
closely at lynx news access for a while) would probably be something
like

 * use LIST to get details of available groups, if the URL includes a 
   wildcard, otherwise just try to access the group and see if it exists
   (GROUP command). [Loading an entire active file over dialup just to
   read a single group would be irritating...]

 * use XGTITLE <pattern> if supported by the server, to get just details
   for the group(s) requested (e.g. just comp.* if the URL specified that)
   or if not supported then

 * use LIST NEWSGROUPS if available to get a complete list, else no 
   descriptions available.

The code would need to keep track of whether a full active file was on hand, 
or just details of explicitly named groups, and where descriptions were not
available it would need to distinguish the cases of description sought but
not available and description not yet sought (e.g. LIST done, so full group
list available, but user only viewed comp.* so no descriptions retrieved yet
for other groups). Depending on when lynx needs descriptions available, you 
might also need to use XGTITLE to get extra descriptions other than those 
related directly to the user request (e.g. for groups mentioned in 
cross-posted articles).

                                John Line
-- 
University of Cambridge WWW manager account (usually John Line) 
Send general WWW-related enquiries to address@hidden
;
; To UNSUBSCRIBE:  Send a mail message to address@hidden
;                  with "unsubscribe lynx-dev" (without the
;                  quotation marks) on a line by itself.
;

reply via email to

[Prev in Thread] Current Thread [Next in Thread]