geoff@utcsstat.UUCP (Geoff Collyer) (06/04/84)
I'm going to comment further on Dave Anderson's remarks. I said The attitude that "Gee my VAX has a 6 megabyte data space, so no one needs to ever think about memory constraints" is widespread and dangerous. Dave said Very true. But it is the progressive view, and I'd rather push the limits than be forced to hang back. Don't confuse ``popular'' with ``progressive''. If the progressive view is dangerous (ruinous, not risky) then I'll be regressive, thanks. I said It is possible to fill even a 6 megabyte data space. The advantage of facing memory constraints squarely is that your programs have a good chance of running on even quite small machines. Portable programs can not assume a 6 megabyte data space. Dave said Why not? Paging allows an address space as big as the moon. If you worrying about the physical length of an address, well, I grieve for you. That's why we got rid of our PDPs. Wrong (unless ``the moon'' is pretty tiny). Paging allows a finite and sometimes surprisingly small address space. I believe that in 4.1BSD and 4.2BSD as delivered that limit is 6Mb of data. I can't speak for Uglix (system n). No matter what that limit is on your machine, failing to design programs to deal with limited memory is often due to sheer laziness and is often the mark of an amateur programmer. (I realise that this is not always the case.) Portable programs must attempt to run on machines with small memories since by definition if a program doesn't run on small machines it is significantly less portable than those programs that do run. I said B news is not wonderfully portable ... some will say that if expanding news limits breaks PDP-11s ... it's further justification for getting a machine with a bigger virtual address space. There certainly are legitimate needs for bigger virtual address spaces and people hereabouts are looking at machines with same, but in the case of news, I think the problem is poor design. News needs a re-write. Dave said Yeah, news is due for a re-write. Version 3.0 should include screen control, user supergrouping, special bulletin boards, etc. We'll see 2.11 first. But as far as I know, nobody's out there writing it. The design of 2.10 is not poor - it is a product of software evolution. But the real question of virtual addressing will bite you in the end. Sure, you can still do alot without it. But the limitations are too great. Why isn't there anyone using PDP-8's to read news? or 8080's? I think that you ought to at least upgrade to a 32 bit machine. At least while you can still unload your PDP's; Soon nobody may want them. Don't assume you'll see 2.11 news *ever*. I'm going to come out of the closet and admit that I am one of the two people who volunteered to maintain readnews. We have the choice of cramming an elephantine 2.10.1+some readnews into PDP-11s or throwing it out and doing it properly; I hope to throw it out. Damn, man, have you ever *looked* at the 2.10 sources? I keep a special pair of welder's goggles for looking at the sources. 2.10 is not a product of evolution: it's an ugly hack, hacked and rehacked until it can't be hacked no more. I have just written rudimentary inews and readnews in the shell, and each is 38 lines long. So the limitations of memory are *not* too great for a news system. The shell programs are slow and don't do all the things their B news counterparts do, yet, but they demonstrate that address space limits need not be a serious problem for the news programs. This whole discussion about address space size is a bit of a red herring. I'll always want my PDP-11; it's the only UNIX machine I administer that runs v7. Dave said How long till we all need 64 bit addressing? If we start *needing* 64 bit addressing for a measly news system, then computing is doomed.
dman@homxa.UUCP (#D.ANDERSON) (06/05/84)
Ok, Geoff, I give. I still maintain that the pdp 11 is getting to be a bit
of a dinosaur, but hey - I still think they're nice machines and netnews should
work on them. There shouldn't be a division of the net.
Lets see those shells! I understand what you mean by the internals of the
news system; I hacked it to run on our UNIX/370. But I have heard 2.11 rumors
and we have had 2.10.1 for over a year now. If you have a rewrite, please
beta test it and post it. As long as things remain compatable, I think that
most everyone would appreciate better/faster/cleaner code.
As far as 64 byte addressing machines, who knows? If the bio-computists have
there way they'll be building 1 cc memories around 2015, and paging will not
be nearly as important as how to find enough information stick in them.
Dave Anderson 201-949-5552kre@ucbvax.UUCP (Robert Elz) (06/05/84)
Australia doesn't get net.news so we weren't aware that this discussion was happening (until I read it this news item at Berkeley) But, for your info, Michael Rourke, at the University of New South Wales has rewritten 2.10.1 news (essentially entirely) principally in order to make it smaller, and run effectively on pdp-11's. There are several 34's now running news in Aust (though mostly they get no net.all groups as they tend not to have sufficient disk space). The UNSW version is entirely compatible with B news at system interface level, its user interface is slightly different (sometimes needlessly so in my opinion) but not so radically different that you would not feel at home. If you want more information on this, you should send mail to decvax!mulga!michaelr:elecvax (or any of the following equivalents .. decvax!mulga!elecvax!michaelr decvax!mulga!michaelr@elecvax decvax!mulga!michaelr.elecvax Don't post news on this subject to these groups, it won't be seen). Robert Elz decvax!mulga!kre (temporarily kre@ucbmonet.berkeley.arpa)