schwartz@shire.cs.psu.edu (Scott Schwartz) (12/12/88)
In article <14946@mimsy.UUCP>, chris@mimsy (Chris Torek) writes: >For that matter, why do we need object archives in the first place? >They are just a hack to save space (and perhaps, but not necessarily, >time). How about /lib/libc/*.o? Absolutely. Lots of other systems (VM/CMS, for example) have this; people like it, why not try it out under unix. 4.4BSD, anyone? About saving space: isn't it the case that object archives do save space? Less internal fragmentation, and all that? Maybe that is only true under the old filesystem. >(About 1/2 :-) ---the file system is *supposed* to be clean enough and >fast enough to support this sort of thing; why *are* we working against >it?) Hmmm, you might have to stat lots of files looking for the right entry point. I guess we would need a directory oriented version of something like ranlib, in that case. -- Scott Schwartz <schwartz@shire.cs.psu.edu>
guy@auspex.UUCP (Guy Harris) (12/13/88)
>About saving space: isn't it the case that object archives do save >space? Less internal fragmentation, and all that? Maybe that is >only true under the old filesystem. It's true under any file system with a block size greater than 1 byte, which certainly includes the 4.2BSD file system.... >Hmmm, you might have to stat lots of files looking for the right >entry point. I guess we would need a directory oriented version of >something like ranlib, in that case. Open and read, not (just) stat.
andrew@alice.UUCP (Andrew Hume) (12/15/88)
i agitate dfor this in plan9 and ken's response was basically that it would always be too inefficient to do 10 or 20 or 100 or 200 opens to look at object files. with archives, one open is it.
schwartz@shire.cs.psu.edu (Scott Schwartz) (12/16/88)
In article <8542@alice.UUCP>, andrew@alice (Andrew Hume) writes: >i agitate dfor this in plan9 and ken's response was basically that >it would always be too inefficient to do 10 or 20 or 100 or 200 >opens to look at object files. with archives, one open is it. For something like the C library, which gets used constantly, this is sensible, but for libraries that are under development it might be nice to be able to exploit the filesystem. The point is giving users the option shouldn't be too expensive in general. An instance of prior art is X11R3's treatment of fonts: designated directories are filled with font files, and an index file. -- Scott Schwartz <schwartz@shire.cs.psu.edu>