perfsonar development work

["Michael Bischoff" <>]

'Currently' it's envisioned that the psui is to obtain 3 tabs:

- one topology enabled LS tab
(search by topology element which you found by it's geo-location, ala google 
maps)
(or search a service directly by geo-location, if that makes any sense)
This is probably a dumbed down version of a more sophisticated tab that 
circumvents or
'hides' MA/MP and the selection thereof completely from the user.
@man.poznan.pl>, 

 
<>
-'ma.conf' tab (the evolved one I mentioned earlier that functions as 
favorites list. (or
'bookmarks')

- current ls tab (which might be restricted more later on)

As noted before manually selecting  services(mp/ma) by the user might just be 
there as a fall
back method(as in it should only be used in the odd case, or simply because 
the plugin/client
hasn't matured yet) As indicated there's the traceroute approach and I 
suspect more
sophisticated examples in the future. These ways of obtaining services use 
specific queries
which should have results which are limited in size and good candidates for 
optimizations.
This should also improve/remove locality (problems) as people usually measure 
stuff thats
relevant to them(ie close to home)

The reason the gls tab is there in it's form like it is now is because of 
practical reasons:
-It's a demo/test (well it's a little more subtle then that).
-There is a absence of components needed for other forms of use of the 
ls(topology service).
-Testing stuff in isolation(right now we can iron out problems with the ls, 
without being
bothered by topology service - in a similar way we probably want the topology 
service
'tested' by some other form separately before combing them all.

That all being said I still believe that even in the current setup, the 
amount of deployed
services shouldn't scare us or cause fundamental problems. Sure in the end it 
won't scale up,
but I hope by that time psui has evolved as well as the back-end.

This email was originally quite a bit longer but it was getting a bit too 
incoherent. The
meat of it was that the software is a living thing and there also the road in 
between of A
and B. And we have to work on user adoption at the same time(which means 
keeping it usable
while we get there). And at the same time not block or impose on future 
goals. I realize that
it's all bit cliché like.

But yes, we hear you Jeff, your concerns are valid. They have been taken into 
account, we're
just not there yet.

ps. we should probably look into renaming MA.conf - it's doesn't strictly 
contain MA's for
example it also contains MP's and it's hardly 'simply a file' anymore.

-Michael