Path: utzoo!utgpu!news-server.csri.toronto.edu!bonnie.concordia.ca!ccu.umanitoba.ca!herald.usask.ca!alberta!brazeau.ucs.ualberta.ca!unixg.ubc.ca!ubc-cs!uw-beaver!zephyr.ens.tek.com!uunet!alembic!csu From: csu@alembic.acs.com (Dave Mack) Newsgroups: news.software.b Subject: Re: Suggested Improvements to C News Error Handling Message-ID: <1991Jun28.183421.5695@alembic.acs.com> Date: 28 Jun 91 18:34:21 GMT References: <1991Jun26.230703.21677@zoo.toronto.edu> Organization: Alembic Computer Services, McLean VA Lines: 35 In article <1991Jun26.230703.21677@zoo.toronto.edu> henry@zoo.toronto.edu (Henry Spencer) writes: >In article mathew@mantis-consultants.co.uk (mathew) writes: >>IMPROVED ERROR HANDLING FOR C NEWS ... > >Slightly to my surprise :-), this one is not half bad. It needs some >small adjustments to deal with implementation realities -- for example, >you cannot determine the article's date without first doing at least >limited header parsing -- and more thought needs to go into really >outre' failure modes like floods of garbage from insane gateways, but >it's not out of the question. > >The nicest aspect of it, in my opinion, is that it doesn't have to be >done in relaynews at all. Given our existing log entries, plus the >provisions we've been muttering about for another pseudogroup or >something to save certain classes of rejected articles, it can be >implemented as completely separate software that puts log entries >together with articles and posts failure reports. One highly desireable improvement to mathew's scheme (which probably *would* require changes to relaynews) would be to rig things so that the error article is only transmitted to the site from which the original botched article came. Imagine a site connected to uunet which is producing bad articles - it would be much better to have uunet sending the error article back directly rather than transmitting it to every site connected to uunet, and thence the universe. Actually, it might not require a change to relaynews if, instead of actually posting the article, the system simply writes the pathname and size directly into the togo file for that site. togo.error, perhaps? This change wouldn't prevent all the error articles from being sent all over the net, but it might help in some cases. -- Dave Mack