Software Tools Minutes 20 October 1995 9:00 AM - 10:15 AM ====================== Present: Pushpa Bhat, John Hobbs, Alan Jonckheere, Stan Krzywdzinski, Qizhong Li-Demarteau, Lee Lueking, Laura Paterno, Harrison Prosper (from FSU, via a speaker phone), Serban Protopopescu, Silvia Repond Scribe: Stan Krzywdzinski The meeting was devoted to Alan's presentation on the current Code Management, which is going to be a starting point for our discussion on the 3-rd item on our charge list. The written version of the presentation entitled: "D0Library - Organization, Strengths and Weaknesses" was handed out to all present and can be found on D0SFT cluster, in a file usr$root3:[jonckheere.tools]d0library.txt . Limitted number of comments and ideas was exchanged during the presentation, since it was, rightfully so, urged, mainly by Harrison, to forge ahead with the presentation and have the discussion after. The topics of the presentation are listed below, along with the ensuing comments, which should be regarded in context of the presentation text. o What it is - Repository - Archival and code management system John asked whether currently it is possible to retrieve all the code, of a given library, which existed in the past on a certain day, and the answer is yes. o Releases o Strengths of the current system Alan stressed firmly that the main strength of the present system is that it does work ! o Weaknesses of the current system Releases take a long time, the worst case is probably STP library - it takes week(s) to get it out, and the University of Hawai is the last one which gets a release. The official version does not seem to work - there is so much coupling between libraries that linking with the official libraries rarely produces a working executable; an exception is RECO. People have 'libtest all' in their login.com's because they do not know which versions, official or test, of the libraries they need would be compatible. In the future, Harrison proposed to use the version #'s of the libraries, which implies that a developer would have to select a set of compatible versions out of the available pool. John argued that compatibility of the libraries would be secured if a release policy is adopted that every current library becomes an old one at the same time. Regarding beta libraries, John an Serban pointed out that they should have identical structure as the corresponding D0$CMS libraries; John added that the present test and beta versions should actually be replaced by one, say test, version. There are no rules right now on how the beta area should be structured. Laura mentioned prod_db as an example of beta area which is not easily mapped to the D0$CMS, due to a number of different utilities and a usage of SQL precompiler; these require at least 2 level directories, whereas at present CMS allows for one level; CMS with more levels would be easy to implement. Alan remarked that the biggest problem with beta libraries is the code there staying too long. Harrison: compilation on all types of machines / platforms in use should be made automatic upon a developer attempt to release the code. Alan and Harrison were for all the software residing on one huge disk, which could be easily accessible via AFS, enabling cashing of the required parts. AFS does not support multiple versions automatically, but again one may have subdirectories. o Features of "The Next Generation Library" The meeting adjourned at 10:15 AM, with the last item of the presentation, Features of "The Next Generation Library", as well as the full discussion, postponed till the next meeting(s).