Re: You'll never guess what HP advertised

• Previous message: Nic Clews: "Re: Disk errors using ANALYZE or DIAGNOSE"
• In reply to: Karsten Nyblad: "Re: You'll never guess what HP advertised"
• Next in thread: reb: "Re: You'll never guess what HP advertised"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Date: Thu, 13 May 2004 11:45:45 +0200

Karsten Nyblad wrote:
> "Main, Kerry" <kerry.main@hp.com> wrote in message
>
news:FD827B33AB0D9C4E92EACEEFEE2BA2FB3145A9@tayexc19.americas.cpqcorp.net...
>>> True. But most bugs can be fixed without understanding more
>>> than a few
>>> hundred lines of code.
>
>> Please, this is getting silly - you do not fix OS kernel, security,
>> cluster and driver problems by only understanding a few hundred
>> lines of code. Especially ones which deal with loading, cluster
>> interaction, race problems caused by faster servers, intermittent
>> crashes (hw?/os?/isv? software issues?) etc ...
>
> OK, you get a longer answer. I would avoid using an unsupported
> operating system when ever I can. Unfortunately there is always a
> risk ending up running an important application on unsupported
> hardware and software.
>
> Even if you buy operating systems large market share from big software
> vendors, there is a chance that the operating system becomes EOLed
> some time in the future. Just look at VMS. It was a very successful
> operating system for many years. HP might stop supporting it some
> times in the future. Let us assume that happens. There will be
> organizations that will want to use VMS after HP has stopped support.
> What do you think these organization would like if given these two
> options: 1) HP opens the sources of VMS and makes it freely
> available. 2) HP shreds the sources of VMS.
>
> You are right that many of the problems you list, are difficult to
> solve, but do you really think, e.g., security problems are
> necessarily difficult to solve once the problems have been disclosed?
> E.g., do you think you need to understand an entire application to
> fix a buffer over round? In that case I know one consultant I am
> never going to hire.
>

I can name one as well.

> Try thinking about the bugs fixes you have made during your career.
> Do you agree with me in that the main part of those have been fixed
> just a few lines?
>

Some, Yes. Most, no. And in any case, the number of lines fixed
(added/changed/deleted) is not some general function of the complexity of
the problem.

I recall spending 5 days understanding/fixing an 8000 line program to solve
a trivial programming error that required less than 10 lines of code to
correct. Most of the time was spent constructing a logically exhausive test
data set to test all possible inputs in order to isolate under what
circumstances the code produced incorrect results. In other words, figuring
out where in the code to even look. Which category is this in ? This is a
pretty normal state of affairs I think.

Sometimes you have to throw it all away because it is so broken
structurally, that it cannot be patched. Like the classic 80% solution, the
7-10 split, there is no way to reach the goal (10 pins) from the current
position. This is a pretty normal state of affairs I think.

Like someone else wrote, it depends.

However, statistically speaking, the number of people qualified to modify OS
code at client sites is really small and for a host of reasons, I suspect
few sites actually want that responsibility.

Anyhow, this discussion is tangential to the original point. The right and
technical possibility to modify OS code is, in general, a specious argument
in support of an OpenSource OS.

Finally, as an observation on this particular matter,

I have been hearing hotshoe CS grads still wet behind the ears spout this
crap for more than 20 years. Same, Same. Only time and a few spectacular
budget overuns will teaches them any different.

You believe your position to be true, and as a matter of simple psychology,
have invested part of your ego in taking what is technically known as a
*position*. Nothing anyone says will cause you to alter this position, only
*events* which falsify it, and thus create the cognitive dissonance
necessary for a realignment of your mind - and perhaps not even then - are
likely to have affect. Its just the way most people are wired together.
Some minds are sufficiently well trained to avoid the burned fingers
requirement, but they are sadly in the minority.

Dr. Dweeb.

> Karsten Nyblad
> ibpit1202 at sneakemail dot com

• Previous message: Nic Clews: "Re: Disk errors using ANALYZE or DIAGNOSE"
• In reply to: Karsten Nyblad: "Re: You'll never guess what HP advertised"
• Next in thread: reb: "Re: You'll never guess what HP advertised"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]