Re: [ccp4bb]: [cnsbb] Crystallography on OSX

["Warren L. DeLano" <warren@sunesis.com>]

***  For details on how to be removed from this list visit the  ***
***          CCP4 home page http://www.ccp4.ac.uk         ***

On Mon, 22 Oct 2001, James Stroud wrote:

> (snip)
> In summary (from a non-computer science expert standpoint):
> 
>    CNS: Cant make it work with anyone's instructions
>   CCP4: Might work
>    HKL: Nope
> D*Trek: Nope
> Mosflm: No good instructions on making this work anywhere
>      O: Something cool actually works!
> molmol: Might work
>  PyMol: Probably works now, will probably work soon
> 

Regarding hardware-accelerated PyMOL on MacOS X, here is the status: I am
having trouble with a few key issues, such as full mouse support with
Apple's GLUT implementation, construction of a native external GUI, and
lingering stability issues.  The port runs but is only partially 
functional -- it needs more work. 

Unfortunately, PyMOL development for OS X is stalled by lack of adequate
Mac hardware on which to complete the port. 

Thus, for the near term it is likely that only the software-accelerated
GNU/Darwin version will be fully functional.  While that version
works great *NOW* for minor refinement tasks with simple line graphics,
most of the complex representations (surfaces, sticks, cartoon ribbons,
etc.) are far too slow for practical use without hardware acceleration.
Fortunately, ray-tracing and publication-quality graphics work fine with
the current version thanks to Michael Love's impressive work on the
GNU/Darwin port.

Although Apple is quite willing to tout PyMOL as an awesome package for
GNU/Darwin/OS X ( http://www.apple.com/scitech/stories/cornell ), they
have refused to provide essential support for PyMOL development on OS X at
a critical time.  This despite the fact that PyMOL is now in use all over 
the world by hundreds if not thousands of scientists and students
(it will surpass 60,000 hits and 9,000 downloads by week's end).

Given that I am trying to push PyMOL forward on half-a-dozen fronts,
Apple's short-sightedness has convinced me to place the Windows
ActiveX version of PyMOL (a potential open-source Chime replacement) ahead
of the native OS X port in terms of priority.   For practical reasons, I
can't support PyMOL on a platform which I don't have in my development
cluster.

I created PyMOL in my spare time in order to demonstrate the 
remarkable value of completely free and open-source scientific software to
biomedical research.  Since this project is not supported by any
institution, my resources are VERY limited.   If Apple can't even spare a
loaner machine for a few months in order to enable PyMOL to work on OS X,
then I certainly can't spare the time to go substantially out of my way to
support their platform.  Apple doesn't seem to realize that they need
native ports of key software packages far more than we software developers
need MacOS X! 

>From my perspective, the Windows PyMOL ActiveX plug-in will have
substantially more impact than the native OS X port of PyMOL will
anyway.  However, if there is an OS X developer out there with sufficient
time, experience, and current hardware who would like to push the native
PyMOL OS X port to fruition immediately, I am willing to assist.  Likewise  
I would welcome donation (or a 6 month loan) of a 500+ Mhz G4 Mac with an
nVidia card for OS X PyMOL development.

In my opinion, OS X is by far the coolest platform to come along since
Linux, and I hope it will become a serious challenger to XP.  Eventually,
I will even save up the $3000 required to buy a fast Mac and complete the
native port on it (if not already done) -- OS X is far too slick to
ignore in the long term.

However, while I would personally love to be able to develop and support
PyMOL on OS X today, my advice to crystallographers is to avoid the Mac
for serious computing for the time being.  Mac's cost 2-3X as much as
generic x86 Linux boxes and give you less real-world performance than
Intel or AMD systems.  Last time I checked, the high-end Athlon was still
the best bargain for crystallography and computational chemistry.

Compare:  You can now build your own Athlon-based linux cluster node with
half a gig of RAM and a sizable hard disk for around $500, and a
stereo-capable Windows workstation can be constructed for a little over
$1000.  Using a Mac for these purposes would be like paying $1.5-2.5k
extra per machine just to run each copy of the OS.  Sure, MacOS X is good,
but it isn't quite THAT good.

Instead of rushing out to buy a Mac now, give the OS X platform a year to
mature.  Perhaps by then Apple's processors will have caught up with
Athlons & Pentiums in terms of performance, and we developers will have
had a chance to provide native ports which take advantage of the Mac's
unique architecture.  Maybe Apple will even come to the realization that
it is fundamentally a software company and start selling OSX for x86
hardware (now there's some wishful thinking).  

- Warren
warren@delanoscientific.com
http://pymol.sf.net

PS.  My apologies in advance to G."DVD"K. for any spelling, typos, and
grammatical errors which may appear in the above.  Please understand
that my native language is Z80 assembly, and that I tend to have trouble
with words larger than 8 bits.