wunder (03/24/83)
Source is not that much larger than binary, and it is smaller than three or four binary copies. Xlisp (a rather large bit of source) runs on PDP-11's, VAX-11's, and Z-80's. We brought it up on our Onyx (Z8000), with a couple of changes. Rice may bring it up on their SUN (68000). Someone here plans to try it on his 6502. That is six different binaries. They would be submitted from different sites, too, maybe multiple times, and with varying changes to the source. Source code is useful for other things, too. There were a couple of clever tricks in xlisp, like using a #define for a lengthy structure member reference. Of course, a lot of the stuff in net.sources is hideous to look at -- at least I wouldn't be expecting any comments in an object module. There are good text compression schemes that could be used to save some space. The title could warn that the artiles was compressed. Now if somebody would write a program to do that and submit it to net.sources... wunderwood UUCP: fortune!wdl1!wunder ARPA Liaison: wunder@FORD-WDL1
seth (03/24/83)
#R:tekecs:-66400:hp-cvd:8300001:000:787 hp-cvd!seth Mar 21 09:42:00 1983 Will binaries for Berkeley VAX run on System III VAX? How about binaries for 4.1 bsd running on 4.2 bsd? What about 68000 based UNIX systems, or IBM PC UNIXes or LISA UNIXes, and so forth? I suspect that all of the binaries are not interchangeable. So instead of large stuff in net.sources, you have notesfiles for each binary format. Then authors of software in net.sources would end up submitting binaries for each group (and chances are they would not even be able to submit software for each group, not having all the different machines and UNIXes). As an alternative, how about splitting up stuff in net.sources. If program foo is large, split it into n pieces which are smaller and more easily transmitted via uucp. --Seth HP Personal Computer Division Corvallis, Oregon
guy (03/24/83)
"Imagine distributing something like UNIX that way?" Well, take a look at the number of micros out there running UNIX, take a look at Ma Bell's pricing structure, and then see how many of them offer binary licenses. Software to be distributed to non-hackers generally is distributed in binary form because half the customers wouldn't know a compiler if it came up and bit them and the other half know, but don't want to use it unless they have to. If you are a software house, you may have other reasons for not sending out source than a desire to keep your lack of talent for programming a secret. It's a *lot* harder to support (a magic word that makes customers a lot happier) software sent out in source form; how do you know what the customer has changed? Trust them? Not likely. Compiles take longer than links or copies, are more complicated, and are more error-prone. Programs that can be reconfigured by changing some text file are a lot easier for naive users - and even experienced users! - to change than programs where you have to change the source and recompile to reconfigure them. Ever done an RSX-11M system generation? Ever done a USG UNIX system generation? I've done both (well, a PWB/UNIX system generation, anyway); to do the former, which involves a reassembly and relink, you have to set aside the better part of a day (and there, at least, DEC's system generation software, in principle, does all the changing of the source for your), while to do the latter takes about 15-30 minutes - and only about 5 minutes of that is editing the configuration file. No, I don't think that source is the obvious method of choice for distributing software. Sometimes it's better than binary, sometimes it's worse. Remember, the average computer user, the average system administrator, even the average system programmer is not as much a clever hacker as the average UNIX system programmer. On the other hand, if you restrict the considerations to what is sent out over USENET, "net.binary" is not very useful. You can't run VAX-11 binaries on a PDP-11, and you can run PDP-11 binaries on a VAX-11 only if it doesn't use separate I/D space, and even then it's not the best way of doing things. And if you aren't running a 11-family machine, neither PDP-11 nor VAX-11 binaries are much use. I suspect the average net.sources subscriber is clever enough to deal with source distributions. Guy Harris RLG Corporation seismo!rlgvax!guy
jsq (03/26/83)
Count me for another no to net.binary.
kimf (03/27/83)
Here's another "no" for "net.binary". The Pioneer ogcvax!kimf