From: jklowden@speakeasy.net   
      
   On Wed, 15 Feb 2012 22:21:42 -0800 (PST)   
   "MikeWhy"  wrote:   
      
   > > You seem to be saying that 250,000 lines of code relying on any   
   > > class X can't easily be changed to use class Y, for logistical   
   > > reasons.   
   >   
   > I'm saying a project that has homegrown Vector likely has homegrown   
   > unique_ptr, lock_guard, and other resource wrappers   
      
   I see.  I'm coming at it from the opposite direction, assuming to the   
   contrary that code that avoids std::vector is unlikely to be   
   sophisticated about e.g. pointer sharing and locks.   
      
   > The task is exponential with proliferation of all   
   > X's that you replace with standard or other Y's.   
      
   Harder, yes, exponential, no.  I think you meant the term dramatically,   
   though, right?  You don't mean that for some effort E the work grows   
   with the number of classes N as E^N.  I'd say linear, not exponential.   
   To the extent that a bunch of tightly-coupled classes are all   
   eliminated at once, less than linear.   
      
   > > Yes, non-functional change for the sake of standardization or   
   > > library adoption is work.  But it's not particularly hard, let alone   
   > > impossible, not least because the problem is well specified.   
   >   
   > I don't believe you for even a   
   > moment that it is "well specified." [...] Even speaking only of   
   > vector<>, do you also look for and fix every usage that allocates   
   > vector-like constructs?   
      
   Hmm.  By well specified, I meant that the project of standardization   
   involves no functional change.  It must behave after exactly as it did   
   before (modulo performance considerations).  I'm always more concerned   
   about receding goalposts than the number/kind of internal changes.   
   After all, you really *can* say (or not) that all references to a   
   particular structure, class, or built-in are to be eliminated.  And I   
   don't see the difficulty in specifying what's meant by e.g.   
   "vector-like"; there aren't that many variations on a theme.   
      
   --jkl   
      
      
   --   
         [ See http://www.gotw.ca/resources/clcm.htm for info about ]   
         [ comp.lang.c++.moderated.    First time posters: Do this! ]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)