blarson@dianne.usc.edu (bob larson) (08/27/90)
(I have sent a much more detailed version of this message directly to c-news@utstat.toronto.edu) New C news administrators should be warned that the first run of expire will change the size of history.pag in a radical manner. (Also true any time the history.pag is recreated.) I spent a fair amount of time tracking down a non-bug. I've modified batcher not to include Xref: headers. Actually quite simple. If there is demand, I'll post an unoffical patch here, but would prefer that it just gets incorperated into the next offical patch. -- Bob Larson (blars) blarson@usc.edu usc!blarson Hiding differences does not make them go away. Accepting differences makes them unimportant.
henry@zoo.toronto.edu (Henry Spencer) (08/28/90)
In article <26817@usc.edu> blarson@dianne.usc.edu (bob larson) writes: >I've modified batcher not to include Xref: headers. Actually quite >simple. If there is demand, I'll post an unoffical patch here, but >would prefer that it just gets incorperated into the next offical >patch. Unless you've found a really clever way of doing it, it's not likely, I'm afraid. You can't just use egrep, because whatever you do has to affect headers *only*. As mentioned in notebook/rfcerrata, we consider the don't-send-Xref requirement to be erroneous, since no news system we're aware of -- including B2.xx, which the RFC is supposed to be describing -- implements it. We'd be willing to comply with it but only if it was essentially free, and imposing significant processing in the batching isn't. -- Committees do harm merely by existing. | Henry Spencer at U of Toronto Zoology -Freeman Dyson | henry@zoo.toronto.edu utzoo!henry
flee@dictionopolis.cs.psu.edu (Felix Lee) (08/28/90)
Actually, it's pretty cheap to get rid of Xref: in "batcher", if you assume that Xref: lines are always first. It's somewhat less than the cost of a strncmp plus a strchr per article. Doesn't really feel like it's worth it though, given that it's just as easy to ignore Xref: on the other side. -- Felix Lee flee@cs.psu.edu
blarson@dianne.usc.edu (bob larson) (08/28/90)
In article <1990Aug27.172424.18636@zoo.toronto.edu> henry@zoo.toronto.edu (Henry Spencer) writes: >In article <26817@usc.edu> blarson@dianne.usc.edu (bob larson) writes: >>I've modified batcher not to include Xref: headers. Actually quite >>simple. If there is demand, I'll post an unoffical patch here, but >>would prefer that it just gets incorperated into the next offical >>patch. > >Unless you've found a really clever way of doing it, it's not likely, >I'm afraid. You can't just use egrep, because whatever you do has to >affect headers *only*. It's a change to batcher.c, and the only clever part about it is noting that C news puts the Xref: line first. If the first 6 characters of the file are "Xref: " the portion before the first newline is not put in the batch. (and the size on the "#! rnews" line changes to match.) A fair portion of the code is parinoia about Xref lines that exceed the 8192 byte buffer. (It would be simpler to impelment a policy decision to not send such articles on, but the batcher is not the place to put policy decisions.) -- Bob Larson (blars) blarson@usc.edu usc!blarson Hiding differences does not make them go away. Accepting differences makes them unimportant.
henry@zoo.toronto.edu (Henry Spencer) (08/29/90)
In article <26827@usc.edu> blarson@dianne.usc.edu (bob larson) writes: >It's a change to batcher.c, and the only clever part about it is noting >that C news puts the Xref: line first. If the first 6 characters of >the file are "Xref: " the portion before the first newline is not put >in the batch... Hmm. That's a reasonable approach, indeed about the only reasonable one. I am still reluctant to incorporate even a modest lump of code to deal with a requirement that (a) every other news system ignores, and (b) we think should be dealt with by fixing the specs. Are there really many people troubled by this? -- TCP/IP: handling tomorrow's loads today| Henry Spencer at U of Toronto Zoology OSI: handling yesterday's loads someday| henry@zoo.toronto.edu utzoo!henry
geoff@athena.mit.edu (Geoff Collyer) (08/29/90)
bob larson:
> ... noting that C news puts the Xref: line first.
This is *not* guaranteed, by code or documentation. It is quite
possible for relaynews to emit Xref: as other than the first header,
so writing code that relies upon Xref: being the first header is
unwise. (Consider an article with an enormous header which triggers
Plan B of article filing: since the newsgroups aren't known for
certain when the first headers are written to disk, an Xref: cannot be
emitted before writing the start of the headers to disk.)
Geoff Collyer, wishing that the Cornell-Toronto IP connection would
hurry up and speed up.