[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: EW Community Center, POSIX compliance comments



I certainly support Mitch Withers' comment that splinter groups are
inefficient, and that the earthworm community should volunteer to get needed
jobs done without relying on an overtasked Earthworm Central (EC).  And I agree
that EC simply has too much on their plate to always be responsive to the
entire earthworm community.

Perhaps a better way is needed to facilitate the community's efforts to help
themselves.  There seem to be many capable volunteers writing code, carrying
out community tasks, and contributing useful additions to earthworm.  But it
also seems that some of that work goes unexamined, unrecognized, and unused.
 Mitch Withers' request for a CSS/SEGY formatter, and Kent Lindquist's response
that he wrote one in 1995 and submitted it to EC (with updates), seems to
confirm this (or at least confirms the need to use the list more often).

Given the over-extension of EC, we waste scarce resources if the needed work of
capable community volunteers is "lost" before it can be incorporated into the
distribution.  Perhaps it's just a problem managing resources.  If the
community is to substantially contribute to the development of the code, and if
a single group is to maintain the distribution, then there needs to be an
effective mechanism to get the community's enhancements into the distribution
in a timely way.  Steve Malone's idea for a Community Center manager at EC who
would be responsible for fostering community development of the code is worth
pursuing.  Perhaps the EAB could also look at establishing processes and
standards for handling the community's contributions.

The POSIX version of earthworm is not meant to be an exercise in splinter
groups.  Just the opposite.  It is to allow earthworm to run on additional UNIX
platforms without requiring a separate code base, and without requiring USGS to
support anything beyond the NT and Solaris platforms.  The idea is to replace
the proprietary Solaris API with the industry-standard (and Solaris-supported)
POSIX API.  Developing (or advocating) standards-based code for open source is
usually a "good thing" - and not necessarily the sole domain of disaffected
splinter groups.  The same can be said about the need to address computer
security issues.

The POSIX effort was also not done in a vacuum.  It began after meetings with
EC and their managers last year.  Since we were the ones needing earthworm on a
non-Solaris platform (SGI) we did the work (in keeping with the self-help
principle).  Providing NEIC and others with our data was a significant
motivation for us to run earthworm.  In return, we asked the EC to incorporate
the changes so we would not have to maintain a splinter version of the code.
 EC's help was needed to validate the code, and to provide access to their
production environment to ensure that the POSIX version under Solaris
functioned properly.  Again, we expected to do much of that work.

The (alpha) version of the POSIX code was released in April, even though not
fully implemented on Solaris (and certainly not ready for production).  This
was done to share progress with other developers (and to get feedback,
suggestions, or patches from those who wanted to dive in).  Also, it appeared
that there would be some delay before EC would be able to work with us, and I
didn't want to go much further until they could.

The state of the POSIX code today is pretty much where it was in April:  it
runs on SGI, but compiler flags, etc. aren't yet set up for POSIX under
Solaris. Security issues have also not been thoroughly examined, but the code
appears to be vulnerable to certain spoofing, DoS, and other attacks.  This
applies to both the distribution and POSIX versions.  Several people have
worked on porting to Linux.  With the release of version 5, I'm hopeful that EC
may have some breathing room to work with us and others on a POSIX compliant
earthworm, and that this effort can be completed fairly quickly.



-- 

 Chris Wood	Geophysicist
		USBR Seismotectonics & Geophysics Group, Denver CO
		Internet:  chris@seismo.usbr.gov
		Voice:  (303) 445-3187
From: aln@ccr.jussieu.fr (Alexandre NERCESSIAN)
Message-Id: <200006091118.e59BIxt57824@moka.ccr.jussieu.fr>
To: earthw-list@nmt.edu
Subject: POSIX ew

I reported as soon as 19 or 20 th of april my success in installing
the Chris Wood's port on Linux Intel machine (directly to Chris and not
to the list). Since it seems to interrest more people I report it to
the list now.

Concerning the discussion of the last days, I can understand how difficult
it is to keep coherency in the developpement of such a rich collection
of codes as ew.  For ad hoc purpose, I wrote and modified several modules
but in a way which is not "ew clean" and wich are not ready to be
released to the communauty. Amoung the things which are of interrest for me:
Physical parametrization of the picker (removing hard coded references
to 100hz sampling rate).

Another back end to the eqproc suite (more adequate to volcanic networks
where the difference of altitude is of the same order as the distance
between stations).  All that requires a lot of testing and once it works,
it is useless without a proper documentation which is not the easiest part.

Alexander Nercessian IPGP Paris France
aln@ccr.jussieu.fr
nerces@ipgp.jussieu.fr


----- End Included Message -----