Re: replacing dbm_* with dba_*

On Sun, Jun 20, 1999 at 10:49:41AM -0400, Jim Jagielski wrote: > Sascha Schumann wrote: > > > > > > > > The one thing I seem to recall (I'll check later today and tomorrow) > > > is that dba_open did not support the 'c' open (r/w but no truncate > > > if the file exists) > > > > 'w'. Please look at the documentation. > > > > 'r' - read > > 'w' - read/write > > 'n' - truncate, read/write > > There's another posting I just send out about that... :) > > > > > > and ./configure didn't correctly determine when using db1.x and > > > not db2.x. > > > > We don't support db1 (if you mean Sleepycat's DB1.xx) > > Hmmmm.... since dbm_ does, then I think dropping dbm_ for dba_ > will cause compatibility problems. Some people make not even > be aware that they are using ndbm ala db mode. I'm not sure which interface DB1 uses. Maybe the code supports it by accident? :-) The include files of glibc-2.1.1 in db1/db.h seem to imply that the API is comparable to DB2. I'll thankfully accept pointers to more documentation. > Does not having a write-only mode make sense? That is RDWR without > CREAT? Again, dbm_ implements 'w' this way. Does having a write-only mode make sense? > Basically, these are just issues involving dropping dbm_ and using > dba_. There are differences between the 2. Not sure if it makes sense > to adjust dba_ to be like dbm_ and thus be able to remove dbm_ > (or map it internally to the dba_ function)... > > That's for you guys to decide... :) If you'd like me to work a patch > that addresses these, then I will, but then it means that dba_open's > 'w' behavior will change :/ Changing APIs is a Bad Thing(tm). Especially, if it's not absolutely necessary. -- Regards, Sascha Schumann Consultant