smith@darwin (Steven Smith) (02/21/90)
Can someone tell me why MIT insists on shipping source only? It seems to make little sense to ship 40meg of source so that people can generate 12meg of installed software (manual and all) after six hours of disk grinding. Wouldn't it make much more sense to reduce net traffic and unneeded recom- pilation by having binary tar files for the x different arch's (or at least the most used ones?). I appreciate how much effort went into the imake makefile generation so that one source tree will compile on all machines. But it seems that a simple binary+library shipment is all that most people want anyway. Just my humble opinions. Steve Smith smith@origin.life.uiuc.edu -- Steve Smith Research Programmer Center for Prokaryote Genome Analysis University of Illinois
jkl@eplrx7.uucp (Jay Llewellyn) (02/22/90)
From article <1990Feb21.150131.10382@ux1.cso.uiuc.edu>, by smith@darwin (Steven Smith): > Can someone tell me why MIT insists on shipping source only? It seems to > ... > But it seems that a simple binary+library shipment is all that most people > want anyway. > > Just my humble opinions. > > Steve Smith > smith@origin.life.uiuc.edu > -- Why should MIT be responsible for doing all that work? They're not in the computer business; if people get only binary+library then they are going to expect someone to support that product, MIT, I don't think so. If you read this group, some people have problems getting X up and running. Do you think that MIT could create binarys thay would work on every machine in every possible environment? No flames intended. jkl -- -- Jay Llewellyn | E.I. Du Pont de Nemours & Co. eplrx7!jkl@uunet.uu.net | Engineering Physics Laboratory (302) 695-8209/7395 | P.O. Box 80357 "Maybe it was the roses" | Wilmington, DE 19880-0357 -- The UUCP Mailer
jg@max.crl.dec.com (Jim Gettys) (02/22/90)
For what machines? For what versions of what operating systems? Back in early X10 days we did for a while. Probability of them working was not good, and caused more trouble than it was worth. X now runs on more different machines, and more different releases of operating systems, than you can possibly imagine. Think of the cross product of machine types, operating systems, and releases of base system software, not to even mention different display types. This is what commercial vendors do for you; produce binaries and fix problems. It is much more work than you can imagine doing the testing and distribution. They earn their money, from my experience. For a small group of people doing interesting work that runs on many different systems, shipping source is the only way to go. You want just binaries; go to your computer vendor. That's what they are in buisness for. Or you can go into buisness yourself, selling binary distributions... - Jim Gettys
bob@MorningStar.Com (Bob Sutterfield) (02/22/90)
There are many and varied differences between machines, and maintaining a collection of up-to-date binaries for even a few is a major task. The Consortium certainly can't afford to have one each of all the software and hardware permutations in-house. For a community software effort it's important for everyone to have the sources, so everyone can contribute, and feels encouraged to. Many times, I'm not sure a behavior is a bug (or even if I've seen it correctly) until I examine the source. Surviving the rigors of the building and installation process often turns an X neophyte into someone who can provide at least a little on-site support. (This isn't so true with R4, but in the good old days... :-) It's much easier to hide Trojan horses in a binary than in source.
rick@hanauma.stanford.edu (Richard Ottolini) (02/22/90)
In article <1990Feb21.150131.10382@ux1.cso.uiuc.edu> smith@darwin (Steven Smith) writes: >Can someone tell me why MIT insists on shipping source only? It seems to Agreed! How about stashing binaries for any workstation company with >10% market share?
dinah@NICOLLE.BCM.TMC.EDU (Dinah Anderson) (02/22/90)
Integrated Computing Systems (ICS) will send you binaries for most common
(Watch me get in trouble now...) systems like Sun,etc. for a real reasonable
price. i believe the brochure I just got said $450. That is worth the price
of not compiling it yourself!
I have done business with these guys before and they do good work. They
also offer support and consultation services. They can be contacted at
(617)547-0510 or info@ics.com.
Dinah Anderson Manager of Systems Integration
Baylor College of Medicine Houston, Texas
internet: dinah@bcm.tmc.edu uucp: {rutgers,mailrus}!bcm!dinahkaleb@mars.jpl.nasa.gov (Kaleb Keithley) (02/22/90)
In article <470@med.Stanford.EDU> rick@hanauma.UUCP (Richard Ottolini) writes: >In article <1990Feb21.150131.10382@ux1.cso.uiuc.edu> smith@darwin (Steven Smith) writes: >>Can someone tell me why MIT insists on shipping source only? It seems to >Agreed! How about stashing binaries for any workstation company with >10% >market share? And who would those be? What criteria should we use; some companys biased opinion of where they are in the marketplace? Better yet, why don't some of the sites who have compiled R4 make the binaries available via anon ftp. (In the spirit of X Consortium, why don't the member companies lead off by making the binaries for their own hardware available first.) Chewey, get us outta here! kaleb@mars.jpl.nasa.gov Jet Propeller Labs Kaleb Keithley
rws@EXPO.LCS.MIT.EDU (Bob Scheifler) (02/22/90)
I told myself to ignore this message. I told myself to ignore this message.
Can someone tell me why MIT insists on shipping source only?
I must say, I am continually amazed at the willingness of the public to
demand arbitrary amounts of work from other people.
It seems to
make little sense to ship 40meg of source so that people can generate 12meg
of installed software (manual and all) after six hours of disk grinding.
Haven't you learned yet that X is a vendor conspiracy to sell more memory and
disks?
Wouldn't it make much more sense to reduce net traffic and unneeded recom-
pilation by having binary tar files for the x different arch's (or at least
the most used ones?).
Why don't you and yours make your own binaries available in this fashion?
But it seems that a simple binary+library shipment is all that most people
want anyway.
Probably more than 70% of the bug reports we initially received were from
people who ran into problems trying to build different configurations than
what we ever tested (much less ran on a normal basis) at MIT. Different
OS versions, different compilers, different destination directories, different
almost everything. Binary distributions would be of no help to those people.
And of course, every so often we put out source patches; I suppose you want us
to put out complete new binary distributions with each patch as well? And
no doubt you're willing to give me some money to pay for the extra disk space
we'll need, and the extra person we'll probably need just to try and keep all
those binaries up to date; I'd like unmarked $20's in used suitcases, please.smith@darwin (Steven Smith) (02/22/90)
It seems that some have misunderstood my intentions from my original posting. I do not advocate removing the source, simply including pre-compiled versions of a few flavors (sun3,sun4,MicroVax,DecStation,hp,sgi,etc.) under the most resent release of their OS's. As far as trojan horses are concerned, I doubt that ANYONE searched through the entire source tree before simply trying a 'make World' on their copy of R4. And I would be much more likely to trust MIT for a clean release than any other site offering such a service. I have my source in tack, and I do go to it quite often. But that still doesn't explain why one must have it to use X (Thats a lot of code!). I would think that MIT would be more than happy to lower net traffic. And it would be nice if it didn't take a full day to get the software running. For those of you that are interested, the University of Illinois has the pre compiled code available for the Sun3 (3.5 and 4.0.3), and the Sun4(4.0.3c) on ux1.cso.uiuc.edu (128.174.5.59). I, they, and everyone else make no claim to its condition with respect to viri. But I have been using it for a month now. This is a service intended for the use of on campus sites, but I doubt that they would mind if others used it. More opinions, Flame me directly, support me publicly, Steve Smith smith@origin.life.uiuc.edu -- Steve Smith Research Programmer Center for Prokaryote Genome Analysis University of Illinois
scs@iti.org (Steve Simmons) (02/22/90)
rick@hanauma.stanford.edu (Richard Ottolini) writes: >In article <1990Feb21.150131.10382@ux1.cso.uiuc.edu> smith@darwin (Steven Smith) writes: >>Can someone tell me why MIT insists on shipping source only? It seems to >Agreed! How about stashing binaries for any workstation company with >10% >market share? Sounds reasonable. Lets see, since no single workstation/operating system has 10% of the market (and none is likely to in the near future) . . .
fischer@iesd.auc.dk (Lars P. Fischer) (02/23/90)
In article <1990Feb21.150131.10382@ux1.cso.uiuc.edu> smith@darwin (Steven Smith) writes: >Can someone tell me why MIT insists on shipping source only? It seems to >make little sense to ship 40meg of source so that people can generate 12meg >of installed software (manual and all) after six hours of disk >grinding. There simply are too many different types of machines. And even for one architecture, some people like /usr/lib/X11, others prefer /home/local/X11/lib, and so on. Also, remember that everytime there is a patch, The X Consortium, or MIT, would to recompile the whole thing for N different machines. Too much work. >But it seems that a simple binary+library shipment is all that most people >want anyway. If there really is a need for this, some large Sun site, say, could easily make available a SPARC binary dist on some machine. Sun.com would be a good choice :-). Remember, you get what you pay for ... /Lars -- Lars Fischer, fischer@iesd.auc.dk | One thing the UNIX system does not need CS Dept., Univ. of Aalborg, DENMARK. | is more features. -- Kernighan & Pike