Re: Re: "Drivers? We don't need no stinking..."

Well, I do seem to have hit a nerve.  I don't like being called a "flamer",
but my intention was to raise the issue for serious discussion, and that
seems to have happened, so I can't complain too much.

My attempts at subtle humor can be taken as abrasive and confrontational
sometimes as well, but sometimes that's the most efficient way to get things
going.


> Alan wrote:
> 
>>>> The are documented drivers not ported from 3.3.6. There are
>>>> documented drivers with bugs. here are unaccelerated documented
>>>> chipsets. There is full documentation for old sis 3d, for trident
>>>> 3d and there is broken code for sis3xx 3D.
> 
> Both of us were discussing the issue of devices/drivers that were supported
> by XFree86 3.3.6 and are still not supported by 4.x. I was explaining that
> documentation exists for everything that was supported by 3.3.6, that I had
> a copy of the vast majority of them, and that David Dawes now has them. I
> also indicated that doing this work (porting old drivers to 4.x) was a
> really good way to help address the "dearth of mentally competent
> programmers" as you so tactfully put it.

I inferred that that must be what Alan was talking about, when he said lack
of documentation was not the only factor, for the following reasons.

First, he didn't specify what other factors there might be, so I was left to
infer.  Second, Alan is closely associated with Linus Torvalds, who in recent
times has delivered scathing attacks on programmers who have submitted code
to him, suggesting they must be using drugs.

Third, you yourself publicly belittled the efforts of Mike Harris, offering
that as a reason why he wasn't granted commit privilege.

What the @#$% else was I supposed to infer?  In my long experience, XFree86
has always billed itself as free software, written and maitained by
volunteers.  Why would it occur to me that money might be a issue...?!?


> So show me exactly where I said anything about issues of keeping up with
> current hardware.

You didn't, but I did.  Show me where my original post said anything about 
obsolete hardware.


>> Neither addresses my, uh, treatise on bloatware, which 
>> implies a problem with future maintainability of code.
> 
> You didn't write a treatise. You wrote a flame, to which Alan responded by
> pointing out that there is still plenty of hardware with plenty of
> documentation for which documentation exists for which drivers need to be
> written/ported/fixed.

Of course I didn't write a treatise.  That's self-deprecating humor, believe
it or not.  As for the rest, see the preceding.

 
>> With the Wretched Offender's 'kdrive' as an example, perhaps 
>> it is possible to prune the code.  As for chips, maybe I
>> should replace my aging G400 and 300 MHz K6-2 with a
>> 2 GHz CPU or two, and run dumb framebuffer mode; that was
>> Intel's suggestion, several years ago.
> 
> Vote with your dollars. If you are committed to the Open Source-only path,
> the don't by hardware from vendors that don't release docs. This has worked
> in the past, rarely.

Which only helps make my case about the need for a lawsuit.  Not that I
expect there to be one, mind you, but I do think that properly litigated one
would be successful, and that's why there needs to be an enlightened dialog 
on the issues that would be involved in one.


>> For the future of XFree86, then, my two main concerns are the 
>> same as form certain other "large" Open Source projects: 
>> ongoing hardware support, and sustainability of the code base.
>> I have chronically parted company with the prevailing views on
>> how to deal with both of these issues, throughout my 9 
>> years running Open Source systems:
>> 
>>   1. On hardware:  A lawsuit should have been pursued years 
>>      ago, already. With the finding that Microsoft is an
>>      illegal monopoly, that would be much easier now.  The
>>      prevailing argument against this ignores the "reasonableness
>>      clause", such that the legal remedy cannot impose an
>>      unreasonable burden on the defendant in a suit. 
> 
> You're kidding, right? There is no correlation between any finding w.r.t.
> Microsoft and intellectual property rights protection. Somehow Microsoft
> having been found guilty in its case has implications on whether hardware
> vendors are allowed to choose their own paths for protecting their
> intellectual property? Sigh...

Register usage is not intellectual property.  Legally, it is an "essential
facility", which any driver writer/maintainer needs access to.

Chip vendors don't need to betray any IP in register-level docs; the IP is 
inside the circuitry inside the chip, and chip vendors routinely betray more 
about that in glossy promotional materials than they ever could in register 
docs.

Register usage is legally equivalent to a network protocol or a printer 
protocol.  You must have reasonable documented access to it to fairly use 
the hardware.

This is a market manipulation issue, not an IP issue.  The lines between
conflicting legal rights have not been legally drawn, and until they have 
been the situation cannot be rationalized.  (I avoid the word "stabilized", 
as that seems a tad laughable in the market at issue.)

Microsoft enters the picture because, as a monopoly, it is both reasonable
and healthy that people should seek (and seek to provide) alternatives.  But
when they do, they come in direct competition with Microsoft, even as
individuals competing for control of their own personal machines and Internet
activities.  And in order to compete effectively, people must have access to 
the hardware.

It has been shown in abundance that this cannot be left to the hardware
vendors to decide arbitrarily for themselves.


>>   2. On the code:  Separate components are preferable to 
>>      monolithic systems (Linux does this, of course). 
>>      For each component, render a basic version extremely lean
>>      and mean, with more elaborate and ambitious alternatives
>>      available as options, using separate but coordinated code
>>      bases (God help me if I ever have to put my code where 
>>      my mouth is on this one, but I am serious). 
> 
> And the issue with the XFree86 4.x loadable modules architecture is exactly
> what? The entire point of the loadable architecture was to enable vendors to
> choose their own IP protection policies and release their own drivers if
> they have a business case to do so. Unless you're an "intellectual property
> is evil" zealot, this should be good enough.

First of all, I was talking about the whole ball of X wax, not just the 
server; and as source, not just as binary.  The server modularization scheme 
has only increased its bloat, not reduced it.

But that was meant to be a separate discussion, anyway.  That's why I have 
two separate points numbered above.  Apparently it wasn't separate, for you.

So about modules, WHO says it should be "good enough"?  What if I think the
available driver is crap, and want to try writing a better one?  What if the
supplied driver has a distracting bug?

What if I want to run something other than X?  (Now please don't accuse me of
flaming for that....)



-- Jim Howard  <[email protected]>