Minutes for 28 June 1996 Tools Working Group Meeting
----------------------------------------------------

Present: Herb,Laura,Lee,Qizhong,Serban,Jerry,Stan,Alan

Topics: 1. Configuration Management report 
	2. DataBase evaluation 
	3. Library structure proposal

Minutes:

Configuration Management report (Alan)
--------------------------------------

 Alan presented a summary of what was presented by Don Petravick in 
yesterday's 
CD Run II meeting. A document representing the current reccommendations for 
code management can be found on the Web as D0->comp->softwaretools->CMGT->
DraftDoc (last). This is updated on a continual basis. The main difference
between the approach taken for run II and the current system is we would like
to use mostly canned components and only write the "glue" between them. Such
components which seem to fill our needs are CVS,gnumake and UPS/UPD. Other 
specific tools like cpp are being discussed. One desire is that code be 
compiled when checked in, as a zero level verification; this will be possible.
It is felt that that no requirements be made for testing tools and bug 
tracking.

Alan spent some time outlining a possible tree structure for the CVS as the 
following brief example:

		
		     da top
		     ------
			|
	---------------------------------
	|		|		|
     ------- 	     -------	    ---------
     package  	     reco	     utility
     ------- 	     -------	    ---------
     | | | |         | | | |	    |   | | |
   ----		    -- - - -	  -----
   YBOS	            V1 2 3 4      EMACS

(I hate drawing these pictures, so don't blame me if you cannot figure it out!
ask Alan for more details)


DataBase evaluation (Laura)
----------------------------

Laura presented an overview of what was presented in Boston plus some 
additional information which was not. Currently we use 4 or 5 data base systems
and would like to avoid this in the future. There was some discussion about
whether we needed centeralized or distributed databases as well as what this 
really meant. Laura summarized the cost of the current system in terms of $ and
FTEs to maintain. She showed price quotes for Object Store, Objectivity and 
Versant, roughly as follows for UNIX (this table is GREATLY simplified for 
clarity, for more info talk with Laura or Jerry) :

			Object Store	Objectivity	Versant
	--------------------------------------------------------
	Developer	$12K		$7.8K		$6.7K
	Runtime server  $6.6K		$2.5K?		$14K
	runtime client  $1.2K		$150		$630
	--------------------------------------------------------
	
Objectivity visited last week and described their product. Some advantages 
seem to be 1. scalability, 2. Platform independence, 3. lots more. One concern
is over the way the platform dependent byte structure is dealt with and this
issue requires more understanding. CERN has chosen Objectivity due to its 
scalability both in terms of large data sets and large number of users. It is
the opinion of the DB group that OODBMS are the way to go. More evaluations
of the top candidates are planned for the coming weeks.

Library Structure Proposal (Serban)
-----------------------------------
Serban has distributed a proposal for the run 2 library structure. After some 
discussion over how lists might be used to define libraries I was a little
confused. It is clearly tied to the configuration management structure, and 
the proposal was to generate separate libraries for each major production 
exe , eg RECO. We decided that it would be useful to attempt to develop the 
tree structure which Alan started (above) a bit more and see how things seem 
to fit together. Qizhong  asked Serban if he could do this for the  next 
meeting.  


Faithfully submitted,


Lee