carlson@uxh.cso.uiuc.edu (02/19/89)
One question on name servers: I have got the Berkeley Domain server daemon (BIND) named to run, and the debugging output indicates that it is receiving info from higher-up servers, but none of the programs seem to be able to access the daemon. Is it that the gethostbyname(3) call needs to be recompiled for named, or have I overlooked something? Another reason for source code--I may not want to modify it, but looking at it will tell me the actions needed to correct a problem. Also, gethostbyname, etc, those library routines and the utility programs that depend on it could be recompiled WITHOUT AFFECTING DEVELOPERS' APPLICATIONS AT ALL! In this kind of a case, I may want to make a minor and transparent change (by obtaining the proper network routines), but it would require a large amount of source code and recompiling to install the change. (Of course, this would be simpler if Unix had dynanamic linking.)
guy@auspex.UUCP (Guy Harris) (02/21/89)
>Is it that the gethostbyname(3) call needs to be recompiled for named, Possibly. I think nExt uses Yellow Pages, which means access to named would normally go through the YP server. >(Of course, this would be simpler if Unix had dynanamic linking.) SunOS 4.0 has it, and the 4.0 mechanism will be the basis for the System V Release 4.0 mechanism. Sun has already used it to replace "gethost*" with versions that talk directly to "named". I seem to remember somebody from NexT telling me that NeXt has dynamic linking as well.
tytso@athena.mit.edu (Theodore Y. Tso) (02/21/89)
In article <72100003@uxh.cso.uiuc.edu> carlson@uxh.cso.uiuc.edu writes: > >One question on name servers: >I have got the Berkeley Domain server daemon (BIND) named to run, >and the debugging output indicates that it is receiving info from >higher-up servers, but none of the programs seem to be able to >access the daemon. >Is it that the gethostbyname(3) call needs to be recompiled for named, >or have I overlooked something? Without source code, you cannot use bind without sufferring with/under Yellow Pages. Not having source code, and not having the time to check, I'm not sure, but that probably means that you have to also suffer from all of the security holes in Yellow Pages, but life's tough. Grrrr. Yet another example of losing utterly w/o source code. And universities are supposed to provide support?!? Anyway, the trick is to start ypserv as ypserv -i, so that the YP will forward hostname resolutions to named. Unfortunately, that means that you have to play the entire YP game, and that means you have to build YP database files for /etc/services, /etc/passwd, /etc/group, and all sorts of other files. =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= Theodore Ts'o mit-eddie!mit-athena!tytso 3 Ames St., Cambridge, MA 02139 tytso@athena.mit.edu If it's for real, it isn't!
pvo1478@neptune.uucp (Paul O'Neill) (02/21/89)
In article <9360@bloom-beacon.MIT.EDU> tytso@athena.mit.edu (Theodore Y. Tso) writes: > >Anyway, the trick is to start ypserv as ypserv -i, so that the YP will >forward hostname resolutions to named. Unfortunately, that means that >you have to play the entire YP game, .......................... Actually, you can play 1/2 the YP game. Don't put the `+' in /etc/passwd that enables users from the YP database. Now, if you have an existing YP server, you get all the name-service goodies from it by running ypbind on your NeXT. You don't get all the yppasswd security bugs. Just about the only down side I can think of is you'll be dead-in-the-water whenever your server crashes. Moto: Let someone who knows what they're doing manage the nameserver. You get all the services "for free." I ran ypbind on our cube for the first couple days it was here. All the nameservice and mail-alias stuff seemed to work great. The boss said to turn it off, though, cuz' the NeXT docs said it would interfere with other applications. Has anyone run 0.8's ypbind for extended periods. Is it up to snuff? (Frankly, it seems to crash just as often now as it did when ypbind was running :-) Paul O'Neill pvo@oce.orst.edu Coastal Imaging Lab OSU--Oceanography Corvallis, OR 97331 503-754-3251
jackson@adobe.COM (Curtis Jackson) (02/22/89)
In article <72100003@uxh.cso.uiuc.edu> carlson@uxh.cso.uiuc.edu writes: }WITHOUT AFFECTING DEVELOPERS' APPLICATIONS AT ALL! }In this kind of a case, I may want to make a minor and transparent change You do, of course, understand what a cliche' the above is. It ranks right up there with "the check is in the mail" and "it'll only hurt a little bit." The first time I heard it was at college when someone made a "minor and transparent change" (those *exact* words) to the early C compiler on our machine. Afterwards, the following program: main() { int i; i = 5; printf("%d\n", i); } produced a large floating-point number as output... This is not to imply you are that incompetent, but merely to raise your awareness of what the phrase means to those who have been bitten. -- Curtis Jackson @ Adobe Systems in Mountain View, CA (415-962-4905) Internet: jackson@adobe.com uucp: ...!{apple|decwrl|sun}!adobe!jackson