On Wednesday 05 October 2005 21:19, Devdas Bhagat wrote:
On 05/10/05 17:25 +0530, Kenneth Gonsalves wrote:
On Wednesday 05 Oct 2005 5:08 pm, Devdas Bhagat wrote:
Go ahead. Software developers will issue warranties only in cases where they know and control the entire state of program execution. You can't complain if a car refuses to run under water and it was never advertised as doing so. If it fails under a set of controlled conditions, the manufacturer is liable. Multitasking operating systems just aren't controlled environments.
spose i gave a warrantee like: warranteed to work with
os distro x version y apache version x python version x wxpython version x
No. You have to guarantee exactly what is running on the CPU. And you need the exact same model of CPU. If you CPU switches registers too early, too late, has a few cycles more, the guarantee is void. You would want something like kernel version x.y.z, glibc a.b.c, apache version x with compile time flags C,O,M,P.I,L,E running with priority X and every file matching MD5 hashes...., and running on CPU <foo> over motherboard <bar> with hard disk model M from vendor v with firmware version F.v, etc running no other programs (including interactive shells)...
and that the defect complained off should be demonstrated on my box with the above setup? (this is just a rough idea, to be refined before we issue the GPW)
This is about what you get currently. If a bug can be reproduced, it can be fixed. Just no formal warranties.
You do get exactly the above - a formal warranty for fixing bugs that can be reproduced by a well defined set of conditions - when you purchase hardware & software from good vendors.And you fix the cpu-ram-mobo-whatever problem by making customised products defined to do just what u specified. People who need such things - petrochem, pharma industry, defence, medical electronics - pay fat sums for the privelege. Now try fixing your problem running some closed program on your defined-to-the-last-solder-joint-hardware. Even M$ gives you the source for windows CE - even a mad man wont touch it otherwise. Which only goes to show how much faith they have in the drivel they dish out in their ads and PR.Libre software makes no assumptions and tells u point blank. I dont see a problem with that. The author of the bbc article is cunning (or clueless ?), while pretending to take issue with all software merrily goes on to make idiotic pronouncements about libre software which truthfully states facts. Someone gifts u a pen and says it's great for signing. You sign a check without enough dough in the bank and get clobbered in court. So sue the pen gifter for inducing you to sign checks without educating you in law, commerce, economics etc. fater all he told u it's great for signing.
rgds jtd
On Thursday 06 Oct 2005 9:32 am, sherlock@vsnl.com wrote:
You do get exactly the above - a formal warranty for fixing bugs that can be reproduced by a well defined set of conditions - when you purchase hardware & software from good vendors.And you fix the cpu-ram-mobo-whatever problem by making customised products defined to do just what u specified. People who need such things - petrochem, pharma industry, defence, medical electronics - pay fat sums for the privelege.
ahh. this is what i mean. Why cant a quality conscious team developing a foss app aspire to this? Just because some companies give crappy warranties that warranty nothing it doesnt mean that open source developers should follow suit.
ppl would be willing to pay fat sums to us too for the priveledge. Certainly do-able and should not be done with too many nitpicking specifications. The extent and nature of liability accepted also has to be carefully specified with relation to the nature of the software package in question. Car manufacturers do exclude liability for breakage of glass parts, headlights etc.
-
Kenneth Gonsalves -
sherlock＠vsnl.com