cwr@pnet01.cts.com (Will Rose) (03/20/90)
In a recent message charles@hp-sdd.hp.com (Charles Keith) writes: >I was interested, but not concerned. Now, having patched (but not compiled) >1.5.5, I am not only concerned, but a little upset. There seem to be two >problems at work here: >1. The upgrade process is long, tedious and error prone. Once you are out >of sync or have lost a file, endless time can be wasted trying to recover. >If you get out of sync and lose the path back, the only recourse is to >whine on the net (could somebody send me...), wasting bandwidth. I don't think you can upgrade above 1.3 without a hard drive, unless you are a hero (and have a certificate to prove it - a death certificate). If you have a hard drive, surely you backed up the old system before messing with it? No? >2. The upgrade postings were riddled with defects and errors. Each problem >requires diagnosis and manual attention, and sometimes fiddling of the most >desperate variety (perhaps if I add this line the crc will come out right). Riddled is a bit harsh - I've upgraded from 1.3 -> 1.5.3 without checking crcs (a complicated story) and when I came to sort them out, with the help of Bruce Evan's checkcrc, I found fifteen [.sxhc] files off. The problems were in whitespace, and I corrected all but four of them. These four had been posted by ast without uuencoding, so the message system had munged them and their patches to the umpteenth generation. Crc's seem to be nice but not essential if postings are uuencoded (and if they aren't uuencoded, crcs are worthless). I personally thought for something containing 4 MB of data, the 1.3 -> 1.5 upgrade was pretty slick. > Seriously, this kind >of stuff is driving people away from minix, and soon only the captive audience >(academics and students) will be left. Nope. >The proposed solution was to identify and register Real Licensees and give >them an easier way to upgrade. But unless administration suddenly becomes >free, this will not work (think about the government). You want a free operating system, you gotta pay for it (Sam Goldwyn.) >should all shout a suggestion in Andy's ear: TEST! Promise an undergraduate >some extra credit and lock him/her up in a room with the old version and >some upgrade postings. I've long been amazed that ast doesn't do this - I work at a US University where undergraduates are considered an expendable resource; posssibly the fine democratic traditions of Amsterdam prevent this. Distributing unchecked patches is anyway a very poor example to students studying the implementation of a major software system; and I'm serious here. >Problems with shar'ed binaries and empty files should never get out of the >Netherlands. Yup. >2. Lift the ban on posting certain parts of the system in their entirety. What about net volume (apart from 1.3 -> 1.5.0), and what about PH's legal department? >3. Create (graduate project) an automated upgrade system to compliment the >automated posting system. This must be robust and well tested, and try to >anticipate problems that would torpedo the process in midstream. I'll drink to that. Will ======================================================================== The danger of the road is not | in the distance, | cwr@pnet01.cts.com Ten yards is far enough to | break a wheel. | Meng Chiao. | ========================================================================= UUCP: {nosc ucsd hplabs!hp-sdd}!crash!pnet01!cwr ARPA: crash!pnet01!cwr@nosc.mil INET: cwr@pnet01.cts.com