wohler@spam.istc.sri.com (Bill Wohler) (05/04/88)
i sent the following to a group of people as a way of letting off some steam. a couple of them, who work at hp, suggested i post it here for the benefit of the hp engineers. and so now, uncut, comes the cries of one engineer who would rather be doing something else... ----- Forwarded message follows ---- Date: Tue, 03 May 88 11:22:59 PDT In-reply-to: Dave Kumpf's message of Tue, 03 May 88 10:16:23 PDT. <8805031717.AA05201@ucdavis.ucdavis.edu> Subject: Re: sanitized unix hp-ux boring dave? hardly. now that our software has to run on it, we have been spending all of our waking hours porting software to it. it's very challenging coming up with filenaming conventions to get around the 14 character filename limit. rather than go out to the movies, we watch our files with .c, .ms, .1 and ,q suffixes all get merged into one file. it's very amusing reading all the neat stuff in the man pages only to see the disclaimer at the bottom of each saying that it isn't supported on the 300s (which is what the army, and thus us, uses). it's fun doing dumps when you can't partition the disk. when you do full dumps, you have to dump /usr and all that other neat system software that never changes (we put all the static software in a read-only /usr, dump it once, and ignore it for a while). we're a networking shop and so we do lots of networking. on other systems, the operating system does not allow more connections if inetd has used up all available file descriptors. not hp-uxs. inetd accepts more connections than it can handle, chokes and then dies so no more connections can be made. you can't even get an ICMP echo into the kernel! this is fun stuff! oh i could go on! you even have to ifdef your .cshrc. no more /usr/ucb. vi is in /usr/bin! thanks for csh and vi, bill joy, but we're not going to create a /usr/ucb! imagine how stressful it must be to live in israel. at any moment your car could blow up. a user on an hp client lives in the same environment. if the server reboots, all of its clients reboot also. poof! so much for your editor session on that file. so much for your graphics program that has been running for two and a half days. why not a stateless client model like the sun? and all because hp has good marketers. how did they talk the army into accepting hp as a standard? unfortunately, we developers deal with the problems and produce running software, so our managers and hence our clients see everything running all nicely and so they are happy with their decision. they would, however, have seen much more software if their standard had been something more reasonable. (i won't mention any names.) i haven't even begun to learn where system v has put all of the /etc files. the only thing that gives me any hope is that hp is quite responsive to user input, unlike some areas in sun. well, thank you. that session has made me feel much better. --b "i'm going away to tahiti" w
raveling@vaxa.isi.edu (Paul Raveling) (05/05/88)
In article <12409@sri-spam.istc.sri.com> wohler@spam.istc.sri.com (Bill Wohler) writes: > > i sent the following to a group of people as a way of letting off > some steam. a couple of them, who work at hp, suggested i post it > here for the benefit of the hp engineers. > ... I think it would be appropriate to place some blame where blame is REALLY due: At the roots of Unix. HP has done ALMOST as good a job of producing a maximally compatible Unix port as is possible. The biggest exception is those blasted 14-character filenames -- HP deserves a bloody black eye for all the trouble they've caused that could have been avoided. Speaking from experience with HP-UX, BSD, System V, Version 6, and non-Unices (VMS, TOPS-10/20, several PDP-11 systems [including EPOS], OS-360, PC-DOS, + maybe a dozen more), I'd say the general rule is: If any given software entity ports easily from one Unix to another, it's either a miracle or a consequence of somone's blood, sweat, and tears. In fact, I've found it easier to port at least a few programs from PC-DOS to Unix than from one Unix to another. The reason lies in Unix's architecture: It originated with a simple kernel, which produced a fairly high ratio of capability to implementation effort. The overall system relied on ability to assemble tiny pieces into useful entities, particularly via shell programming. It was a hackers' system, that provided the tools for everyone to 'do it yourself'. They did. This applies not just to different breeds of Unix -- it also applies to trying to keep a common software environment in order for use by every member of a project or an organization. Anyway, this is just part of the reason why I used to keep a "Unix Sucks" slogan in my xload window. People demanded something less offensive due to all the demos we do on my machine, so now it says "Give Me Liberty or Give Me Unix!". Anyone want to fund development of a system called Liberty? --------------------- Paul Raveling Raveling@vaxa.isi.edu
ajs@hpfcdc.HP.COM (Alan Silverstein) (05/10/88)
> ...post it here for the benefit of the hp engineers. We are, at least, listening. > The biggest exception is those blasted 14-character filenames... The Series 800 2.0 release already supports long file names. About the Series 300 I can say nothing at this time.