Thu, 20 Jun 1996 04:31:24 +0200
|
On Wed, 19 Jun 1996 21:14:49 EDT Roger Fajman <[log in to unmask]> said:
>Greg had the fun job of bringing up LISTSERV on Solaris and handling the
>Unix side of the conversion from VM LISTSERV without having used
>LISTSERV before and with very little documentation.
I can certainly see the difficulty. Migrating a list base of NIH's size
is no easy task to begin with, even for a LISTSERV expert. In general the
most successful "migration stories" have been from sites like VT,
American or Syracuse (this isn't a comprehensive list so don't kill me if
you're not listed :-) ) where there was significant local expertise for
both source and target systems. I've seen people who know everything
about LISTSERV and VM run into serious trouble because they were not
familiar with the target system, and conversely unix wizards who can find
errors in the sendmail books can make mistakes due to lack of VM
knowledge. NIH also had the misfortune of starting a full migration (ie
with no option to keep "just a few lists" on VM for a while longer)
before the file server or database functions were available, even as a
beta. I am soon going to migrate SUNET's lists from SEARN to SEGATE and,
while my list owners are not the vociferous kind (with one notable
exception ;-) ), I can certainly see that having a database search
function plus the ability to let them manage their own archives of
academic papers is making them a lot nicer than they might otherwise have
been. This is why I am pushing for releasing 1.8c ASAP and keeping all
the enhancements that are still on the "to do" list for later. Many
people will be migrating soon and I can see first hand what kind of
difference this makes in list owner response. A poll I had conducted
about a year ago wasn't quite what I would call positive :-) I have also
carefully pointed out to the list owners that they would gain the web
archive interface (which you can preview at http://segate.sunet.se - I
know that the image link is bad and it's on purpose) and most of them
appear to see the move as a positive thing.
Another category of "migration successes" is people who keep the lists on
VM and migrate the deliveries elsewhere, because it's so much simpler
technically and operationally. SJU is the perfect example, they were #10
worldwide before their migration in SEP95, and now they are #2 (yes, they
have overtaken UB!) All right, now the 4381 is starting to run out of
steam just forwarding the deliveries to the AlphaServer, but it's a small
machine and it's the #2 worldwide LISTSERV site too. The main advantage
of this approach is that on the day the traffic is migrated there is
pretty much no visible change to the users, other than mail arriving much
faster. For a successful traffic migration you don't need to know much
about VM or LISTSERV, in fact what you need to know is how to tune the
mail system on the target machine and install/configure a basic LISTSERV
system with no list and no user. It's a *lot* easier than a full
migration and can buy you time during which L-Soft will continue to port
the remaining VM functions. I don't work for SJU but I do know a few of
their list owners ;-), and I think there would have been a minor natural
disaster in their general area if they had decided to migrate all the
lists to VMS last summer :-) Now it just might work, especially if they
can keep a small number of lists on VM for a while.
Eric
|
|
|