lyndon@auvax.uucp (Lyndon Nerenberg) (01/01/89)
In article <1988Dec30.012400.18278@ziebmef.uucp> cks@ziebmef.UUCP (Chris Siebenmann) writes: > Unfortunately, Smail 3.1 is trying to replace sendmail, not smail 2.5 >and is of an appropriately large size. I'm planning to stick to smail >2.5 unless smail 3.1 gets a lot smaller. In what way is smail too large? It's running quite happily on a few diskless 3/50's here (plus just about every other machine in the shop). The binaries come in at roughly 250KB. That's not even *close* to the size of our emacs binaries. It's also HALF the size of the paths file it has to work with :-) The source tree is getting close to 4 megs, but you don't have to keep that online. I wouldn't count on smail 3.1 getting any smaller -- there are still additions to be made (e.g. DECNET support). One nice thing is that the code is very modular, therefore the binary you generate doesn't have much fluff in it for capabilities you don't need or want. -- Lyndon Nerenberg Computing Services Athabasca University {alberta, attvcr, ncc}!auvax!lyndon || lyndon@nexus.ca
cks@ziebmef.uucp (Chris Siebenmann) (01/02/89)
In article <428@aurora.auvax.uucp> lyndon@auvax.UUCP (Lyndon Nerenberg) writes: >In what way is smail too large? It's running quite happily on a few >diskless 3/50's here (plus just about every other machine in the shop). >The binaries come in at roughly 250KB. That's not even *close* to the >size of our emacs binaries. It's also HALF the size of the paths file >it has to work with :-) Smail 2.5 here is 19K (using a shared library and stripped), and the paths file (I use a smart host) is 1K. When you only have 67M of disk space, such considerations matter. Since I'm a uucp-only site, all I really need is the bugs in smail2.5 fixed. >One nice thing is that the code is very modular, therefore the binary >you generate doesn't have much fluff in it for capabilities you don't >need or want. How big is the binary for a uucp-only site? Somehow, I suspect the small simple sites of the world have been forgotten in the quest for ever more neat features. -- "Hey, thats's why Dad moved to Colorado. He sits Zen now. Me, I like the mountains." Chris Siebenmann uunet!{utgpu!moore,attcan!telly}!ziebmef!cks cks@ziebmef.UUCP or .....!utgpu!{,ontmoh!,ncrcan!brambo!}cks
dave@lsuc.uucp (David Sherman) (01/03/89)
In article <1989Jan2.001754.11708@ziebmef.uucp> cks@ziebmef.UUCP (Chris Siebenmann) writes: >In article <428@aurora.auvax.uucp> lyndon@auvax.UUCP (Lyndon Nerenberg) writes: >>In what way is smail too large? It's running quite happily on a few >>diskless 3/50's here (plus just about every other machine in the shop). >>The binaries come in at roughly 250KB. That's not even *close* to the >>size of our emacs binaries. It's also HALF the size of the paths file >>it has to work with :-) > > Smail 2.5 here is 19K (using a shared library and stripped), and the >paths file (I use a smart host) is 1K. When you only have 67M of disk >space, such considerations matter. Since I'm a uucp-only site, all I >really need is the bugs in smail2.5 fixed. I agree entirely. This old machine, with its 58 ports, 20-30 users during the day and 50 very active UUCP connections handling about 10Mb/day has 960K of memory, of which over 200K is taken up by the kernel. Back in the Good Old Days, people considered 1Mb a large amount of memory. smail here is 56K. I sure wouldn't like to see it become 250K. (No paging here either -- this is v7, so if you swap out a process it's the whole process.) And no, we can't increase the amount of memory on the machine. David Sherman The Law Society of Upper Canada Toronto -- Moderator, mail.yiddish { uunet!attcan att pyramid!utai utzoo } !lsuc!dave
lyndon@auvax.uucp (Lyndon Nerenberg) (01/04/89)
OK, so it won't run on a HP 41C :-) You can't run emacs there, either,
so I guess it's no great loss.
I'll build a stripped down version and post the sizes. I have a feeling
it will still be bigger than you would like, although I can attribute part
of that to the size of the routines sucked in from later releases of the
system libraries rather than "tons of code providing new features." I
can't recall ever building anything much over 200K on any V7 box :-)
--
Lyndon Nerenberg Computing Services Athabasca University
{alberta, attvcr, ncc}!auvax!lyndon || lyndon@nexus.cabrian@ncrcan.Toronto.NCR.COM (Brian Onn) (01/04/89)
In article <1989Jan2.222329.27995@lsuc.uucp> dave@lsuc.UUCP writes: >[stuff about Smail2.5 being big 'nuff already] >I agree entirely. This old machine, with its 58 ports, >20-30 users during the day and 50 very active UUCP connections >handling about 10Mb/day has 960K of memory, of which over 200K >is taken up by the kernel. Back in the Good Old Days, people >considered 1Mb a large amount of memory. smail here is 56K. >I sure wouldn't like to see it become 250K. (No paging here >either -- this is v7, so if you swap out a process it's the >whole process.) > >And no, we can't increase the amount of memory on the machine. Could I interest you in a Tower? :-) I'll even throw paging in :-) -- +-------------------+--------------------------------------------------------+ | Brian Onn | UUCP:..!{uunet!attcan, watmath!utai}!lsuc!ncrcan!brian | | NCR Canada Ltd. | INTERNET: Brian.Onn@Toronto.NCR.COM | +-------------------+--------------------------------------------------------+
lyndon@auvax.uucp (Lyndon Nerenberg) (01/05/89)
In article <1120@ncrcan.Toronto.NCR.COM> brian@ncrcan.Toronto.NCR.COM (Brian Onn) writes: > >Could I interest you in a Tower? :-) I'll even throw paging in :-) Sounds to me like he already *has* one! :-) I'm still waiting to hear what it is that's so big about smail that makes it not worthwhile running on smaller machines ... [ and yes, I'm still going to post the sizes, as soon as an unnamed technician type around here gets me back on the ethernet :-( ] -- Lyndon Nerenberg Computing Services Athabasca University {alberta, attvcr, ncc}!auvax!lyndon || lyndon@nexus.ca "Ethernet works in practice, but not in theory." -- Dave Boggs
cks@ziebmef.uucp (Chris Siebenmann) (01/10/89)
In article <433@aurora.AthabascaU.CA> lyndon@auvax.UUCP (Lyndon Nerenberg) writes: ... >I'm still waiting to hear what it is that's so big about smail that >makes it not worthwhile running on smaller machines ... It seems to me that smail3.1 will either give me a lot of features I have no use for (like SMTP delivery on a machine with only UUCP), or has got a lot of overhead for the simple and probably most common case. It is highly likely that the aims of the smail3.1 authors do not include making a small effecient mailer for uucp-only systems (or systems with uucp plus an internal network). I have no objection to this; I just wish they would call their mailer something besides smail, so as not to mislead and disappoint all the uucp sites which are expecting something of roughly comparable size and functionality to smail2.5 but with the bugs fixed. smail3.1 sounds far too much like an upgrade to smail2.5; by all accounts, it isn't, so I wish they'd stop confusing people (heck, the smail3.1 list occasionally gets the occasional bug-fix or problem report for smail2.5, a pretty clear indication people are confused about this). >[ and yes, I'm still going to post the sizes, as soon as an unnamed > technician type around here gets me back on the ethernet :-( ] I'm looking forward with great interest to finding out the sizes (if you have a version of smail2.5 handy, compiling it and telling us the comparative sizes would be handy). Oh why can't anyone build small single-purpose programs any more? [Obligatory answer: because smail3.1 is solving a large and intractable problem, not the small and simple one smail2.5 solves. It's just not a problem that I have, so I don't feel I should pay the overhead for it.] [I'll get off my hobby-horse now, I think...] -- "Hey, thats's why Dad moved to Colorado. He sits Zen now. Me, I like the mountains." Chris Siebenmann uunet!{utgpu!moore,attcan!telly}!ziebmef!cks cks@ziebmef.UUCP or .....!utgpu!{,ontmoh!,ncrcan!brambo!}cks