Print

Print


On Tue, 2006-06-13 at 14:37 +0100, Brew, CAJ (Chris) wrote:
> Hi All,
> 
> This came up briefly in the face to face and maybe if we have an agreed
> list we can push we can get something done about them.

My ¤0.02 doesn't cover specific implementation issues, but rather
meta-problems associated with the design and implementation of the
software suite itself.

In no particular order:

Portability.
-----------------------------------------------------------------------
Source code for every component should be portable and readily
available; if I want to build my own binaries or packages for my
preferred flavour of Linux/Unix then it should be easy for me to do.  

RHEL4 and variants aren't supported yet, which by itself is worrying --
but replicating the build process for the entire LCG stack is, just by
itself, really hard. 
   
(Just _finding_ the source code that corresponds to each binary package
is non-trivial; even then, a lot of the source just Doesn't Work Out Of
The Box -- my recent attempt to build VOMS on Solaris is a good
example.[0])


Reinventing the wheel, or The Right Tool for the Right Job.
-----------------------------------------------------------------------
GSI and VOMS do pretty much the same job Kerberos has been doing for a
couple of decades.  

LCAS does pretty much the same thing as PAM has been doing for years.

The conventional domain name system could be used to replace MDS.

The various logical file catalogs do pretty much the same thing as a
BitTorrent tracker.

The various file-transfer tools -- GridFTP, srmcp et al -- seem to do
pretty much the same thing as scp or BitTorrent.

Why develop and deploy all of these grid-specific tools -- many of which
are not mature -- when existing generic tools (many of which are already
well-understood and deployed) already exist?



Okay, so this is only two things.. I'm still thinking about how best to
express the other 3.

Cheers,
David

[0] http://www.gridpp.ac.uk/wiki/DPM-on-Solaris#VOMS
-- 
David McBride <[log in to unmask]>
Department of Computing, Imperial College, London