Conference iosg::all-in-1_v30

Re question 3, there are several topics on this already in this conference.

Re question 5:

What happens to a drawer when the account that owns it is deleted has nothing
to do with the number of "connections" to that drawer, or whether they are local
or remote.  (I put "connections" in inverted commas as I'm not 100% certain what
you mean; I assume you mean a pointer to the drawer in someone's file cabinet.)

If the drawer is private, it is deleted when the account that owns it is
deleted.  This is analogous to a user's file cabinet being deleted in V2.4 when
their account was deleted.

If the drawer is shared, two things can happen to it: either it can be deleted,
or it can be moved to a "holding" account.

If the shared drawer is deleted, other people's file cabinet pointers to it
remain.  So if/when they try and access the drawer, they will get an error
message.  This happens for both remote and local sharers of the drawer.

If the shared drawer is moved to the "holding" account, other people with
pointers to it can carry on using it; they may not even know that it has been
moved or that its owner has gone.  Again, this is the same for both local and
remote sharers of the drawer, as the only change necessary is to update the
PARTITION record on the node where the drawer lives.

The idea of the holding account is just to provide a temporary home for drawers
which contain information which cannot be deleted, but for some reason the
owner's account has to be deleted before a new home can be found for the
drawer.

Ideally, the drawer owner should make arrangements for someone else (one of
the other sharers?) to take possession of the drawer before the account is
deleted: the holding account is meant as a last-chance if that is not possible.

Once a drawer has been placed in the holding account (which by default is the
MANAGER account) the system manager should make arrangements for it to be
moved to a new home.

Scott

    ad 1)
    the dictionaries just have different names, which is obvious because
    they support different languages. Whatever language your customer is
    running, s/he will get the appropriate dictionaries with the kit. All
    they have to do is select it and run the spellchecking.
    On multilingual installations however, you end up with more than one
    dictionary, then you have one for each language installed. But it's
    still the same functionality: select the dictionary you want to use and
    go ahead and spellcheck.
    	Stefan

    arjen,
    
    taking the others...
    
    2) Changes aren't documented in release notes - bugs sorry I mean
       oversights in the design are!! Are you asking about bugs fixed?
       these used to be in a 'differences document' but in  v3.0 are
       incorporated into CM and may be reviewed using the VBC (View Base
       Changes) option.  
    
    4) don't understand this one
    
    6) ALL-IN-1 isn't supported in the configuration you quote - it must
       be 'common-environment' i.e installed and started on just one
       node (this goes for DDS not just the File Cab Server) or all across
       the cluster {every node} in which case the srv73 process uses the name
       of the cluster. I'm sure it can be made to work in a
       'non-common-environment' but it isn't officially supported.
    
    7) i'm not sure but I believe the general advice is to use groups wherever
        possible (then you have one the one ACL other than the owner) but
       there's nothing to stop you having hundreds of ACLs on a directory
       or file though the performance isn't too optimal - there is in fact an
       (in)famous customer in Denver who has a system like this!
    
    Regards,
    
    Andrew.D.wicks
   

    Great guys, thanks for all the quick answers.
    
    regarding 4) Are there any special considerations when upgrading on a
    cluster, i.e. are there special things to look out for when upgrading a
    cluster?
    
    Thanks again for all the help.
    

    Re .0
    
    Question 7: More than 24 is bad, more than 50 is *TERRIBLE*. And I told
    you that in Module SM4 of the ATS in Valbonne - Pay attention :-)
    
    Re .4
    
    Re-install DCL tables on each node, run startup on each node, the
    ususal stuff.
    
    Graham

    Hi Graham,
    
    I'll wash my ears out, I've always had a problem with those.....
    I'm surprised you actually remember me, to be honest, I would have
    thought you would have erased me from memory by now!!!!!
    
    Thanks for the answers though,
    regards,
    Arjen
    

Re: reply 5 - for the benefit of someone not able to attend the seminar in 
Valbonne - could you please give an explanation of why there is poor 
performance with more than 24 & especially more than 50 & what determines those 
numbers.

Pam

    
    VMS can only handle so large an ACL efficiently, when it goes above
    a certain threshold VMS uses the equivalent of continuation records
    to hold the entire ACL, it won't fit in the file header. Once this
    expansion is done, using the ACL becomes alot slower.
    
    --Bob