Leisner.Henr@xerox.com (marty) (08/16/89)
I really don't think improved compression is going to help us significantly. On images (with mostly white space) techniques like RLL encoding can give you 1-2 orders of magnitude of compression. The ascii compression techniques give you a factor of 2 or 3. For C programs (lots of letters, lots of different words floating around), LZW compression (i.e. compress) works pretty good. 16 bits is somewhat better than 13 bits (my experience with large distribution of C code in tar format) is 13 bits will give you a little better than a factor of 2 compression and 16 bit will give a little less than a factor of 3. For distribution, I was unhappy with the tons of compressed ar files -- I'd rather see source trees in compressed tar format. It seems the Minix ar is somewhat different than the other ar's I've used. Tar seems to be inherently portable. Perhaps the answer is not ship all the source code. Ship the binaries, the Minix OS sources and maybe some of the tools source code. There can be a README file about how to obtain the rest of the user contributed software (i.e. uucp, networks, bulletin boards, the guy down the hall, etc.). I don't know how workable this is -- it's just a suggstion (I'd hate to see all the good user contributed software drive up the price of the product). ARPA: leisner.henr@xerox.com GV: leisner.henr NS: leisner:wbst139:xerox UUCP: hplabs!arisia!leisner
ast@cs.vu.nl (Andy Tanenbaum) (08/26/89)
In article <20051@louie.udel.EDU> Leisner.Henr@xerox.com (marty) writes: >For distribution, I was unhappy with the tons of compressed ar files -- I'd >rather see source trees in compressed tar format. For V2.0, I am going to make a single distribution for PC/XT/AT/386, instead of separate ones for the PC and AT. This means that I will have to use 360K diskettes or the PC guys won't be able to read the disks. With such a small medium, tars of whole trees won't work. If somebody wants to hack MINIX-tar into POSIX tar (described in section 10.1.1 of the P1003.1 standard), I'll certainly consider using that instead of MINIX ar, but probably on individual directories, not whole trees in most cases. In any event, I'll include an 'install' shell script next time. >Perhaps the answer is not ship all the source code. One of the advantages of USENET over face-to-face communication is that people can't throw rotten eggs, overripe tomatos, etc. at you. I have a hunch that this suggestion will not increase your popularity in this group. I intend to include all the sources, except maybe for a few very large items that are widely available via other channels, like elle, kermit and the C compiler. >I don't know how workable this is -- it's just a suggstion (I'd hate to see >all the good user contributed software drive up the price of the product). I think that having all the software available in one box is a valuable convenience to many people. Besides, even with 20 disks, I doubt that the price would go up that much (maybe $110 at most). The economics of publishing are funny. The MINIX book is 719 pages and lists for $46. My networks book is 658 pages and lists for $55. Clearly manufacturing cost isn't the only factor. Andy Tanenbaum (ast@cs.vu.nl)
nick_andrew%713.602@fidogate.fido.oz (Nick Andrew) (09/06/89)
Original to: ast@cs.vu.nl
One distribution for all is fine. So is using 360k disks
(even if it does make the package very bulky). Compressed
tarfiles would be a problem though ...
The fact is, installing on hard disk isn't quite the
cut&dried procedure it should be. And compressed tarfiles
(esp. LARGE, multi-volume ones) are useless unless one has a
hard disk on which to un-tar it. I would tend to prefer
distribution in pure compressed form - but each disk
containing a filesystem.
Also, my opinion of the various tools & utilities people
have written & ported to Minix over 2.5 years is that there
is too much to include as part of "base" Minix. In
Australia, Minix 1.2 or 1.1 (?) from Dymocks costs $350.
Another 10 disks would no doubt push this price over $400.
(Australian ... say US$300 is an equivalent ...)
This problem could be alleviated by distributing the "minix
tools" through another source
(Usenet, BBSs, even mail-order or public domain) ... where
people can get what tools they want/need.
Regards to all,
Nick.
--- Zeta
* Origin: Zeta: Unix, Minix, Xenix support (02) 627-4177 (3:713/602)ast@cs.vu.nl (Andy Tanenbaum) (09/07/89)
In article <14646@nswitgould.cs.uts.oz> nick_andrew%713.602@fidogate.fido.oz (Nick Andrew) writes: >I would tend to prefer distribution in pure compressed form - but > each disk containing a filesystem. I agree. I wouldn't consider anything else. However, these file systems might contain tar files. I'll worry about that later. In any event, I'll make a shell script that does the installation. > >Also, my opinion of the various tools & utilities people >have written & ported to Minix over 2.5 years is that there >is too much to include as part of "base" Minix. Should I leave out kermit? zmodem? ELLE? etc? I think most everything that I have floating around now is reasonably good and has a following. >In Australia, Minix 1.2 or 1.1 (?) from Dymocks costs $350. How much does it cost from P-H? I am not sure who Dymocks is, but anybody who wants to can buy MINIX from P-H and then resell it at whatever price they want. P-H has an office in Sydney. Could you call them, find out what the story is, and post that? Andy Tanenbaum P.S. my mail connection to you is nonexistent.