Re: Moderately improved map of Microsoft tying evidence

[Roberto Di Cosmo <Roberto.Di.Cosmo@ens.fr>]

  I strongly support the idea of putting together
  hard factual evidence of the global strategy
  implemented by MS over the years.
  
  What is *very* difficult in this kind of task, though,
  is that software *is* complex, and it takes a lot
  of time to be able to assess the merits and demerits
  of a system in an objective way. I am not surprised
  by all these mails saying "System X is better than system Y:
  I tried both"... If you do not understand the inner working
  of the systems, you can only compare individual experiences
  which can vary wildly. Few people have the knowledge (and the
  time) it takes to point out the origins of problem they faced
  and to assess the seriousness of the design error underlying it.
  
  Actually, this is the very core of the problem: it is too
  easy to state that a given design error is not an error but
  "the best we know how to do in such a rapidly evolving complex
  technological field". To counter such an argument you must 
  explain how to do better to people that will not understand
  what you are talking about.
  
  Nevertheless, I have an example that everybody knows and maybe
  somebody could try to "explain" its underlying technological
  issues in a language accessible to everybody.
  
  I am talking of the defragmentation utilities :-)
  
  I personally met lots of people which were very proud of
  knowing how to use a defragmentation utility under dos/windows
  and were very surprised to see that I had no such magic tool
  on my portable running NextStep (that has a mach based bsd
  unix underneath if you know what I am talking about).
  The typical reaction was: "ah... poor boy has not my magic tool"
  
  I usually take the time to explain that I do not *need* it
  because unix implements a self defragmenting disk management
  policy. Wait, I do not use these words :-) I tell them a little
  story about how would they go about filling in and emptying 
  different sized drawers in their closet to avoid leaving too
  many drawers half-filled. (Dont ask me to do this exercise in
  english: I can get by with what I need for my work as a computer
  scientist, but I am not a novel writer :-)).
  
  Up to now, everybody understood, at the end, that their defragmentation
  utility is not a "feature", but a blatant proof of inferior technology
  and lack of interest in the user's needs.
  
  Here the reactions vary, though: they have been told that a
  defragmentation utility "makes your disk go faster", but not
  that "the silly MS system makes your disk go slower", and
  they do not all handle this information the same way.
  
  
  
  Also in the category "slowing down the technological evolution/fooling the customer",
  we should put the braindead directory scheme used once in DOS and
  modified now for FAT. I mean, unless you use a hammer on your disk,
  you are supposed to never find a damaged directory structure on an
  Unix system (this is guaranteed by the inode data structure + fsck
  algorithm), while I personally have seen many disks completely ruined
  by "ScanDisk" (the last one this summer: ScanDisk "repaired" the disk
  and afterwards all the filenames were turned into gibberish and all
  the data was lost).
  
  Anybody willing to try his hand at writing a little story accessible to everybody
  on these easy to verify facts?
  
  --Roberto Di Cosmo
  
  --------------------------------------------------------------
  LIENS-DMI                   E-mail: dicosmo@dmi.ens.fr
  Ecole Normale Superieure    WWW   : http://www.ens.fr/~dicosmo
  45, Rue d'Ulm               Tel   : ++33-(0)1-44 32 20 40      
  75230 Paris CEDEX 05        Fax   : ++33-(0)1-44 32 20 80       
  FRANCE                      MIME/NextMail accepted                                  
  --------------------------------------------------------------