[Gt-eos] Launching the GridCF

[Oliver Keeble]

Hi,

On 07/11/17 15:52, Frank Scheiner wrote:
> Hi Oliver,
> 
> On 11/07/2017 01:44 PM, Oliver Keeble wrote:
>> [...]
>>> So if we manage to push out a release - even with only minimal 
>>> changes/fixes/updates, maybe just a little polished - before this 
>>> deadline, we would give a bold and assuring statement, that the 
>>> gridcf is taking over (development and) support and users can 
>>> continue to rely on the former known Globus Toolkit - now in the form 
>>> of gct - as technology for their needs.
>>>
>>> I think the earlier we make a sound, the more people can be dragged 
>>> to our cause and the more people can be reassured in keeping the gct 
>>> as technology.
>>
>> This is an interesting comment, as I think it reveals a rather 
>> different perspective from the one we have here at CERN at least. Our 
>> intention is not to encourage people to use gct at all - instead, we 
>> want to keep it working for as long as it takes our community to find 
>> alternatives.
> 
> Aha :-), I think I sensed this on the way but wasn't sure if this is 
> still on the table after so many infrastructures have shown their 
> interest in our cause.
> 
> Why would your community need to find alternatives to gct if "it does 
> the job"?
> 
> Or in general what's wrong with the Globus Toolkit currently that anyone 
> should move away from it - if there's a community behind its successor 
> that will support and maybe further develop it?
> 
> Where would these alternatives come from and what if the people behind 
> it loose the interest in it in the future? Would you not have the same 
> problem at hand then?
> 
> I think we have come to a most interesting point, where we as a 
> community could take the development and support of a core technology 
> used in our infrastructures in our own hands. If this works out, in my 
> eyes this would make sure that we will never run into a similar problem 
> again in the future - unless all of us loose the interest in a core 
> technology of our infrastructures, which is not very likely I believe. :-)

Our community (WLCG) has well over a hundred sites providing resources, 
and we have every interest in making the barrier as low as possible for 
them. We have a profusion of protocols, of which two are not going to go 
away, xrootd and WebDAV. The latter is also our best chance for 
sustainability through technology exchange and for sites who want to 
support multiple (other) user communities. So one protocol (gsiftp) 
fewer is a good thing. This is not to mention the relative difficulty of 
working with the Globus codebase.

X509/GSI has other issues, in particular sending around all-powerful 
proxies and low favour with end users. Here alternative solutions are 
less clear, and indeed there are ways to hide the X509 stuff from users. 
In any case, one can do X509 without Globus dependencies.

As far as I'm aware, we no longer use any of the compute related bits 
(gatekeeper etc).

All of which means we will do our bit to keep this stuff going, because 
we rely on it, but for us this an opportunity to steer in another 
direction. I for one am under no illusions about how long that will take!

Cheers,

Oliver.

> 
> Cheers,
> Frank
> 
> 
> 
> _______________________________________________
> Gt-eos mailing list
> Gt-eos at mailman.egi.eu
> http://mailman.egi.eu/mailman/listinfo/gt-eos
> 

-- 
Oliver Keeble                      Information Technology Department
oliver.keeble at cern.ch                                           CERN
+41 75 411 5965                                    CH-1211 Geneva 23