Yet another fire!? (Was Re: Old stuff in /afs/rhic/opt/brahms)

From: Christian Holm Christensen (cholm@hehi03.nbi.dk)
Date: Fri Oct 19 2001 - 15:02:57 EDT

  • Next message: Djamel Ouerdane: "Use Text_t or Char_t ?"

    Hi Konstantin et al, 
    
    On Thu, 18 Oct 2001 10:18:43 -0700
    Konstantin Olchanski <olchansk@sam.triumf.ca> wrote
    concerning "Re: Old stuff in /afs/rhic/opt/brahms (Was Re: compiling brat/brop/brag in RH7.1 pii computers)":
    > On Thu, Oct 18, 2001 at 12:23:49PM +0200, Christian Holm Christensen wrote:
    > >
    > > Frankly, I'm concerned that people will start using non-standard,
    > > non-supported installations or various third-party software like ROOT,
    > > which you can easily imagine will lead to loads of odd problems.   
    > 
    > You do not want to open this can of worms. 
    
    Too late! And, it's not the first time it's been opened by the way,
    and not always by me! 
    
    > Who decides on what is standard and what is not? 
    
    Well, since Kris, Flemming and I are basically the ones that maintains
    the installations, it's us, in concensous.  Of course this is not the
    dictatorial-trio - we do try to make things as easy to use as
    possible, and we'll gladly listen to complaints, etc..  That said,
    users need to provide compeling arguments (or serious enough trouble)
    for us to change practise, since we've actually given it quite a lot
    of thought, and most of us do have past experience in setting up
    software. 
    
    > How do you keep the standard up to date? 
    
    Currently by hand.  What I'd really like to happen is that we did
    nightly (or weekly) builds of BRAT/BROP/BREG/BRAG/CRASH/... in the
    "new" tree.  Since I have little experience in writting CRON files,
    and also there's the issue of reading AFS passwords from a file which
    hasn't been resolved (that is, it's possible, but do want it?).  Could
    you help out with this? Thanks. 
    
    > How do you keep the standard from changing daily?
    
    Lazyness :-)  No seriously; the "pro" and "old" trees only changes
    when a new minor version of BRAT/BROP/BREG/BRAG/CRASH/... is
    introduced, which does not happen that often (on average at .5
    month^-1), while "new" tree in principel should change with each
    revision update - now, that's not feasible so it basically changes
    when someone (usually Ian :-) requests a new BRAT for use on the
    Farm, which depends a lot on the amount of development going on. 
    
    And also, updates and upgrades are _always_ accompanied with a mail to
    brahms-soft-l so people will no exactly what's going on. 
    
    > And, I do not think you can wiggle out of helping Kris (and others) with
    > their problems by defining their installation as "non-standard".
    
    I'm afraid you read my mail rather quickly.  In fact, I pointed
    numerous solutions, and also that there in fact isn't really the
    problem that Kris anticipated.  
    
    As to "defining their installation ...". I certainly did not mean to
    imply (and I don't think I did, but if I did - I'm sorry) that
    installations outside of /afs/rhic/opt/brahms are bad. It's just that
    the installations in /afs/rhic/opt/brahms are the ones that we use for
    reconstruction on CRS, and I believe also what is more or less used by
    people @ BNL (and other places in the US) not doing development.   
    
    There's a good reason for calling that "standard" (hmm, maybe
    'canonical' or something would have been better), since the version
    numbers are always well-known.  Nothing gets installed in
    /afs/rhic/opt/brahms that doesn't carry a proper unique version
    number.  Hence, it's easy to help people out if they use those
    installations, since you always know which version they refer to.  
    
    The reason I object to having another ROOT installation, other then
    the 'standard' one, lying around on a shared disk, is that it may give
    people the impression that it's actually the 'standard' one (i.e., 
    supported, and the one that BRAT/BROP/CRASH/... is compiled with).
    Thus, to minimize confusion, help us do support, and so on, I'd much
    prefer that users that want a 'special' ROOT installation keep such a
    one in thier ${HOME} directory.  
    
    If the @sys thing of AFS  is something one want's to exploit, we have
    AFS 'home' directories in /afs/rhic/brahms/user/<login name>.  For
    example on my CERN account, which resides on AFS, I have a directory
    structure like 
    
      ~/
      |--.alpha_dux40   -> .alpha_osf20
      |--.alpha_osf20
      |  `-- bin
      |--.alpha_osf32   -> .alpha_osf20
      |--.alpha_osf32c  -> .alpha_osf20
      |--.hp700_ux100   -> .hp700_ux90
      |--.hp700_ux101   -> .hp700_ux90
      |--.hp700_ux90
      |  |-- bin
      |  `-- lib
      |--.hp_ux102      -> .hp700_ux90
      |--.i386_linux22
      |  |-- bin
      |  `-- lib
      |--.i386_linux24
      |  |-- bin
      |  `-- lib
      |--.i386_redhat51 -> .i386_linux22
      |--.i386_redhat62 -> .i386_linux22
      |--.i386_redhat71 -> .i386_linux24
      |--.i386_nt40
      |  `-- bin
      |--.rs_aix32
      |  `-- bin
      |--.rs_aix41      -> .rs_aix32
      |--.rs_aix42      -> .rs_aix32
      |--.sgi_52
      |  `-- bin
      |--.sgi_64        -> .sgi_52
      |--.sgi_65        -> .sgi_52
      |--.sun4m_53
      |  `-- bin
      |--.sun4x_55      -> .sun4m_53
      |--.sun4x_56      -> .sun4m_53
      |--.sun4x_57      -> .sun4m_53
      |-- bin           -> .@sys/bin 
      `-- lib           -> .@sys/lib 
    
    Now that's neat, since it means I needn't care about which
    architecture I'm using.  I just build and install everything with
    prefix=${HOME} and the binaries are put in an architecture specific
    directory - much like when we update ROOT and BRAT, we just set
    prexix=/afs/rhic/opt/brahms/[old|pro|new]. 
    
    So you see, I'm not wiggling out of anything.  I'm just saying: "Help
    us help you, by using the 'standard' installations".  That's really
    the essence.  
    
    > > ... Well, it turns
    > > out, that the fella used something non-standard and therefore you do
    > > not find the problem immediately (spending even more time).  
    > 
    > But it is *your* fault, as a provider of the environment. The reason
    > the user is using something non-standard is always because the
    > "standard way" is not working for him. 
    
    Well, then the user should request a change or help to do what she/he
    needs to do, which is exactly what I did. 
    
    > You can't blame the user when something *you* provide does not work.
    
    I think if you go back and read my mail one more time, you'll see that
    I pointed out, that it does in fact work - not only in principle, but
    also proven by experience. 
     
    > >   7. Release early.  Release often.  And listen to your customers.
    > 
    > So do so. Fix your stuff so people do not have to use non-standard
    > things. 
    
    You're contradicting yourself here.  Well, never mind that.  The
    important point is, that the last part of the maxim stated above, has
    been followed.   
    
    Also, I don't care that people use thier own ROOT installations
    (though I think it's a waste of time and space unless they hack on
    ROOT) if they really want it (and realise that help is harder to get),
    but they should do so in the confienes of thier own ${HOME}, not
    flaunt thier tendencies in general public, thereby sending mixed
    signals.  
    
    Do please read my previous email again.  I think you've misunderstood
    a couple of things, since I truely find your critism at best, far from
    being to the point, and at worst nothing but a flame. 
    
    Yours, 
    
    Christian Holm Christensen -------------------------------------------
    Address: Sankt Hansgade 23, 1. th.           Phone:  (+45) 35 35 96 91 
             DK-2200 Copenhagen N                Cell:   (+45) 28 82 16 23
             Denmark                             Office: (+45) 353  25 305 
    Email:   cholm@nbi.dk                        Web:    www.nbi.dk/~cholm
    



    This archive was generated by hypermail 2b30 : Fri Oct 19 2001 - 15:03:36 EDT