At 04:26 PM 3/6/98 +0000, Mark Landin wrote:
>On Thu, 5 Mar 1998 17:59:13 +0000, Tracy Johnson <[log in to unmask]>
>wrote:
>
>>CM-KSAM is still in CM?  Who would have thought!  :-)
>>
>
>Yeah ... but do they have to be? I mean even a CM KSAM file is still
>just a file, right? Why *couldn't* an CM KSAM file be accessed by NM
>intrinsics?

Sure, it could have been coded to recognize the different KSAM
file types and access both NM and CM type KSAM files.  However,
this would have resulted in a longer code-path and slower overall
performance.  Besides, the NM version combines both key and
data into a single file so a single mapped file pointer can
be used to analyze the blocks and retrieve the entry.  CM has 2
files and as such the same techniques would require 2 mapped
file spaces, different decisions, different limitations, etc.
The files themselves are *completely* different in terms of
their internal structures, etc.


Clearer a design choice for simplicity, performance, ease of
maintenance, etc..  There are probably other issues involved
in the decision as well, which the lab team who did the project
would obviously know better than most.



/jf
                              _\\///_
                             (' o-o ')
___________________________ooOo_( )_OOoo____________________________________

                        Thursday, March 12th

           Today in 1912 - The Girl Scouts were founded.
                    1783 - Washington delivered his "Newburg Address".
                    1933 - FDR broadcast his first "fireside chat".

___________________________________Oooo_____________________________________
                            oooO  (    )
                           (    )  )  /
                            \  (   (_/
                             \_)