OS support and statisitics....

[Gregory Casamento]

Riccardo

On Mon, Jun 14, 2010 at 11:35 AM, Riccardo Mottola <address@hidden> wrote:
> Hi,
>
> first of all: the original question of Quentin was if it is fine to use W2K
> functions by breaking "older" platforms. A clear and sane reasoning showed
> that there is no problem with that, "older" platforms are out of our scope
> or not supportable since mingw does not support them.

Indeed, I realize this.

> This is the topic. We agreed on an answer without a vote against.

Also agreed.  My purpose was to illustrate that while we currently
still support win2k... that we should not go to great lengths to
continue support for it.

> This list has the habit of too often going astray because we are moved by
> personal opinions.  This leads to discussion which is good, but generates
> also often discussion threads that resemble wars and lead to conflicts among
> users and developers.

Yes, it does and we are moved by personal opinions.  This is a
discussion list... personal opinions will continue to be discussed
here for the forseeable future.  That being said, however, if you look
at my original email, I purposefully tried to avoid giving personal
opinions and presented only facts as they stand using the best
statistical data we have at hand to determine usage of certain
operating systems.

While I don't believe the numbers reflected on the website I provided
to be 100% accurate (as I stated in my original email), I do believe
that it does, effectively, provide a "gut feel" for the number of
users who are using a platform overall.

> Gregory Casamento wrote:
>>
>> Win2k is not so common as you might think:
>>
>> http://www.w3schools.com/browsers/browsers_os.asp
>>
>> that data indicates that it's being used by about 0.4% of all users
>> worldwide.   While this might not be the most accurate assessment of
>> the actual installed base, I believe it's the best statistic we have.
>
> As written by David, this kind of statistic is quite unrealistic. Surely,
> few people will use W2K for browsing, but it is still deployed on machine
> like servers or workstations of "Internal" use. SInce we want GNUstep to
> work on Servers too...

David didn't say that the statistics are "unrealistic".. he said they
can be "misleading."  To what degree they are misleading is still a
question in my mind.  Nevertheless, I hardly think that there are so
many installations, such as those described in David's email, that it
would bump the number up above even 1% (if that).

> By the same reasoning it would not be useful to support W2k3 ! Wow, a sloppy
> 1.3%?  Yet go into a datacenter and count up how many servers run W2k3 or old
> W2k...

I have been in quite a few datacenters and there are almost no W2k
servers in any datacenter I walked into and I've been in a few of the
largest in the U.S. (e.g. NASA LaRC, Fannie Mae & First American Title
Co.).     I don't think, however, that any personal anecdotes about
how many there are or are not in any given datacenter are really
relevant since it's not a  source of unbiased information.

> Maybe we should stop supporting Linux? It is used by less than 5% of the
> people browsing.

5% is clearly more than an order of magnitude greater than 0.4%. :)
I, personally, would draw the line at things under one percent (to say
nothing of ONE HALF of one percent), but that's just me.

> GNUstep shall remain of general purpose. Suppose you want to write a server
> application which has a GUI (absolutely standard on a windows system).

No one said it shouldn't remain general purpose, least of all me.  The
point here is that a line must be drawn with respect to what we should
and should not support.

Aside from the argument that there aren't that many users, you may
note that Microsoft itself is dropping support for Windows 2000 next
month as noted here:

http://support.microsoft.com/ph/1131

> Also, our current win32 backend probably almost supports winCE, not a bad
> thing, isn't it?

No, it's not a bad thing.

> A reasonable solution could be to root win32 backend into GDI and leave
> "advanced" support do cairo on windows. And bet, there can be an use of the
> pure win32 backend.
>
>> 3) officially deprecated and will not see either bug fixes or new features
>> 4) will likely disappear in a future release of windows.
>
> GDI has such a large user base and windows relies heavy on compatibility. So
> surely it won't get features, but I don't expect it to disappear so soon. It
> was told dead with Vista yet it still works on 7.

I don't think the reasoning that they "haven't dropped it yet" is
really good enough.   Also I note that you left off two very important
things in the above list.  I will reiterate:

1) issues with compositing
2) not hardware accelerated
(http://blogs.msdn.com/b/greg_schechter/archive/2006/03/10/549310.aspx)

There is no denying that compositing support on GDI is horrible.  The
fact that there is no hardware acceleration only adds insult to
injury.

It may indeed be many years before GDI is dropped, but as you may
note... I listed it's removal as the LAST reason on my list to move to
something else.   The above two are much more immediate and relevant
for us.

> Riccardo

GC
-- 
Gregory Casamento - GNUstep Lead/Principal Consultant, OLC, Inc.
yahoo/skype: greg_casamento, aol: gjcasa
(240)274-9630 (Cell)