jbrown@herron.uucp (Jordan Brown) (05/30/89)
In article <428@qtc.UUCP>, battan@qtc.UUCP (Jim Battan) writes: (Actually, one of the people who responded to him wrote...) > >(or just .bc, or .b-n-c, or .b+c if all known operating systems will > ^^^^ > I know at least one o.s. that will _not_ accept this. Please do not > use this as a name. MS-DOS will hiccup on this and several other names. Unfortunately, there are already newsgroups with names that offend it. (There's something like comp.sys.aux which MSDOS thinks is a reference to the AUXillary device, and comp.lang.c++ which has punctuation that MSDOS doesn't like in a file name.) Of course, it's also offended by names > 8 (or 11 maybe) characters long, but that's easy to translate around. As somebody working on an MSDOS news system (don't ask me for it), I sure wish that all newsgroup names were acceptable MSDOS file names. However, I realize that this isn't the case, and isn't likely. (Besides, due to stupidity in MSDOS, *no* name is guaranteed safe because device names are recognized in all contexts, even with extensions and as directories.) I haven't decided how to handle comp.lang.c++ yet, but I'm assuming that I'll have to handle it somehow. I think it's unlikely that I'll be able to talk people into significantly restricting the punctuation allowed.
loverso@Xylogics.COM (John Robert LoVerso) (06/01/89)
In article <58@herron.uucp> jbrown@jato.jpl.nasa.gov writes: > In article <428@qtc.UUCP>, battan@qtc.UUCP (Jim Battan) writes: > (Actually, one of the people who responded to him wrote...) > > >(or just .bc, or .b-n-c, or .b+c if all known operating systems will > > ^^^^ > > I know at least one o.s. that will _not_ accept this. Please do not > > use this as a name. > > MS-DOS will hiccup on this and several other names. > .... > I haven't decided how to handle comp.lang.c++ yet, but I'm assuming that > I'll have to handle it somehow. I think it's unlikely that I'll be able > to talk people into significantly restricting the punctuation allowed. Newsgroup names shouldn't be bound in any way to the filename restrictions of various operating systems. Just because some other "news" implementations uses such a scheme, doesn't mean that *all* news implementations need to do that, especially those news implementations on some so-called "operating systems" like that "M"-word above. Use a file of newsgroup-to-filename mappings that's used to map from a newsgroup name to a useable filename for the host operating system. -- John Robert LoVerso Xylogics, Inc. 617/272-8140 loverso@Xylogics.COM Annex Terminal Server Development Group encore!xylogics!loverso [formerly of Encore Computer Corp]
wedgingt@udenva.cair.du.edu (Will Edgington) (06/01/89)
In article <2341@xenna.Xylogics.COM> loverso@Xylogics.COM (John Robert LoVerso) writes: ]In article <58@herron.uucp> jbrown@jato.jpl.nasa.gov writes: ]> MS-DOS will hiccup on this and several other names. ]> .... ]> I haven't decided how to handle comp.lang.c++ yet, but I'm assuming that ]> I'll have to handle it somehow. I think it's unlikely that I'll be able ]> to talk people into significantly restricting the punctuation allowed. ] ]Newsgroup names shouldn't be bound in any way to the filename ]restrictions of various operating systems. Just because some other ]"news" implementations uses such a scheme, doesn't mean that *all* ]news implementations need to do that, especially those news ]implementations on some so-called "operating systems" like that ]"M"-word above. ] ]Use a file of newsgroup-to-filename mappings that's used to map ]from a newsgroup name to a useable filename for the host operating ]system. Guess what ? If you and your feed(s) cooperate a bit on this, you don't even have to write any new code. The /usr/lib/aliases file on MSDOS could look like : comp.lang.c++ comp.lang.cplusplu comp.lang.reallylongname comp.lang.unique ... and your feed(s)'s /usr/lib/news/aliases like : comp.lang.cplusplu comp.lang.c++ comp.lang.unique comp.lang.reallylongname The moral: the news software is much more flexible than most people think. -- Will Edgington, Computing and Information Resources, University of Denver System Administrator for udenva (== dueos), dutyche, duorion, dunike, ... {{hplabs,uunet}!ncar,ucbvax!nbires,boulder,cires}!udenva!wedgingt, wedgingt@nike.cair.du.edu, wedgingt@udenva.cair.du.edu, ... (Internet) WEDGINGT@DUCAIR (BITNET), wedgingt@ccndu (CSN/CCN), ... COMING SOON: wedgingt@du.edu (all nets)|"No two addresses are the same ..." 2020 S. Race BA469, Denver CO 80208 | +1 303 871 2081
tneff@bfmny0.UUCP (Tom Neff) (06/01/89)
In article <11756@udenva.cair.du.edu> wedgingt@nike.cair.du.edu.UUCP (Will Edgington) writes: >In article <2341@xenna.Xylogics.COM> loverso@Xylogics.COM (John Robert LoVerso) writes: >]Newsgroup names shouldn't be bound in any way to the filename >]restrictions of various operating systems. Just because some other >]"news" implementations uses such a scheme, doesn't mean that *all* >]news implementations need to do that, especially those news >]implementations on some so-called "operating systems" like that >]"M"-word above. Absolutely right. UNIX in a sense has it "too easy" being able to sub slashes for dots and take it from there. Netnews ought in theory to be runnable on an OS that doesn't even support tree structured directories. (Just don't ask me to write the config file. :-) ) >]Use a file of newsgroup-to-filename mappings that's used to map >]from a newsgroup name to a useable filename for the host operating >]system. Whether you actually drive the mapping off an index file or not is of secondary importance to coming up with an ALGORITHM under each supported OS to map newsgroup names to directory names (or the local equivalent). Because after all, that index file not only has to be *generated* to start with, it has to be *updated* as groups are added. It would be impracticable to expect each news knower to "make up" his own directory names every time someone did a newgroup... and it would hurt consistency net-wide especially in software support. I have no idea what's in the cnews code, but something like int ngdir(char *buf, const char *groupname) with a twisty maze of little #ifdef's would be what I have in mind. The mapping for things like MS-DOS might not be easily reversable, but News should have little need to reverse it anyway. That's what you use the index file for, if anything. >Guess what ? If you and your feed(s) cooperate a bit on this, >you don't even have to write any new code. The /usr/lib/aliases >file on MSDOS could look like : > >comp.lang.c++ comp.lang.cplusplu >comp.lang.reallylongname comp.lang.unique > >... and your feed(s)'s /usr/lib/news/aliases like : > >comp.lang.cplusplu comp.lang.c++ >comp.lang.unique comp.lang.reallylongname This is cute (and would get the job done now for dedicated DOS newsers) but I don't feel as good about requiring lots of sites to reinvent the wheel, as I do about providing some kind of consistent mapping to begin with. -- Tom Neff UUCP: ...!uunet!bfmny0!tneff "Truisms aren't everything." Internet: tneff@bfmny0.UU.NET
loverso@Xylogics.COM (John Robert LoVerso) (06/02/89)
In article <14369@bfmny0.UUCP> tneff@bfmny0.UUCP (Tom Neff) writes: > ...Netnews ought in theory to be > runnable on an OS that doesn't even support tree structured directories. How soon they forget. Back in those long ago, lost ages (the so-called golden era of news), when `A' news was the one and only, newsgroups were all directories at the top level (i.e., no tree structure). John