Dan agreed to our previous plan: when we have a new fixed VOMS server that fixes the problem with the FNAL KCA,we'll give him a copy and he'll upgrade to it.
John asked about Globus 4.0 and VOMS. Here's the deal: the old PRIMA build for Globus 3.2.1 continues to work fine, so we can keep using it. However, it probably won't work when we add real support for x86_64 platforms, so we need to build it there. In addition, it won't work for the web services version of GRAM.
John also asked about GUMS 1.1.1, which he had heard was released to deal with a problem contacting the LCG VOMS. After the meeting, we talked to Gabriele Carcassi, the author of GUMS. He tells us that there is no GUMS 1.1.1, and no problem with the LCG VOMS as far as he knows.
Vikram asked if we could provide a symlink for a header file in OpenSSL. We suggested that he include the file that was symlinked-to instead, and he seemed satisfied with that.
Vikram asked about web services. Alain suggested that he has no experience with GT4 web services, and he might be better off getting support from the Globus Team.
We talked with Vikram about the need for a 64-bit build of PRIMA. He wasn't sure what the issues would be, but promised to look into it. The VDT and NMI teams can provide access to appropriate machines and build environments as necessary.
Vikram was concerned that the installation of PRIMA is not completely automatic, and requires users to move a file into place. We explained that this is because we cannot correctly configure the file because it relies on knowing a GUMS server. If we did move it into place, it would cause users that don't use PRIMA or GUMS to have their installation break until they fix it, and would therefore cause more user support.
Doug asked about a Condor-G problem that he reported. When he changed his executable, it didn't get propagated to the remote gatekeeper. As best we can tell, this is a problem with the GASS cache. Doug is going to try to reproduce it with just the Globus tools. Alain hypothesized that this may be tricky because if the GASS cache is caching based on the complete GASS URL, it will be different for each invocation of the the Globus command-line tools, while it is consistent for Condor-G because it is long-running.
Wayne introduced himself.
Though not discussed in the meeting, here is the rough plan for VDT 1.3.7: We hope to release VDT 1.3.7 in "a few weeks". We're working hard on adding Globus 4 WSGRAM, and we've run into a few problems, but we still have hope that it will be done soon. Major new components in VDT 1.3.7:
With luck, we'll also add support for several platforms that people have been asking for. Some details are available online. Note that this page is currently incomplete, and will be updated as tasks are completed.