Hi,
djmauretto wrote:20% of the time I have not found UNDI, and considering that the computers that I tested were not mine, I could not do the checks deeper.
In any case, what is your problem?
I think (in general) my issue is explained well enough by the words "acceptable reliability".
If "works on 80% of computers" is good enough for your networking, then surely "works on 80% of computers" is good enough for hard disk too. Then you end up with a situation where the OS only works reliably (e.g. both network and hard disk working) on 64% of computers. Then you do video that only works on 80% of computers, and memory management that only works on 80% of computers, and timers that only works on 80% of computers, and scheduling that only works on 80% of computers, and... it doesn't take long before you've got an OS that only actually works on less than 10% of computers. Now if 1000 people try the OS you end up with 100 that think the OS is good. The other 900 people will each tell 10 more people that the OS sucks, and each of those 10 other people will tell 10 more people that they've heard the OS sucks. That's 100 people that think the OS is good and 18900 people that think the OS is bad.
Now consider what happens when you refuse to do things that are dodgy. Programmers tend to be human and make mistakes; so if everything should work (in theory) 100% of the time then everything might only work (in practice) 98% of the time. With 10 different pieces that might work 98% of the time you can calculate "98% * 98% * 98% * 98% * 98% * ...." and you end up with an OS where everything works 81% of time. That is still very bad. For 1000 users; that's 800 people that think the OS is good, and 4200 people that think the OS is bad.
To get sustainable "word of mouth" you need more than half of the people to think the OS is good. To do that, it needs to work on over 99% of computers, and for everything to work on 99% of computers each individual piece needs to work 99.9% of the time. Even if all of your code is perfect (zero bugs) you're going to be struggling to get close to "acceptable reliability" due to hardware/firmware bugs.
djmauretto wrote:Instead of sitting here talking, why do not you do a little test?
Because this approach can never reach "acceptable reliability"; and is therefore a complete waste of time.
Cheers,
Brendan