Mail Archives: opendos/1997/02/05/19:43:52
Once upon a time (on 5 Feb 97 at 14:06) mharris AT blackwidow DOT saultc DOT on. said:
> > on the file system. In fact, DOS wasn't meant to be a multi-FS OS, so no
> > programmer wrote his *DOS* app with that in mind. And you cannot require
> > all the users change their beloved programs (if it's possible at all) just
> > because we changed the root FS.
>
> Since programs that are FS specific don't make sense on other
> FS's then there is no need for them. For example. Norton Disk
Still, you can't expect every user to know that. Remember that programers are
just a small part of the computer users community. What about all those
ladies in banks, post offices, etc? (no offense!)
> Doctor. If you ran that on an ext2 drive, not only would you get
> errors from NDD, but it would't work either. There would be no
> need for NDD either. Unless of course symantec released NU for
> ext2.
You're kidding, right? ;-) Symantec for *free* system? heh,heh - and where's
the money? ;-))
> > DEVICE="C:\System Folder\OpenDOS\Memory Manager"
> > DEVICEHIGH="C:\System Folder\OpenDOS\FAT16 Driver"
>
> ICK!!!!! Filenames with spaces in them are annoying!!!! Also,
> just because there ARE LFN's (or will be), it doesn't mean that
> they SHOULD be used all of the time! ie
>
> C:\BIN, C:\SYS\DOS\FAT, C:\SYS\DOS\MM
>
> These names are good enough. Who want's a 10k PATH? Who wants
You're right, LFNs are good for archive names and documents or directories -
system util/drivers names should be as short as possible.
> > > INSTALL="C:\System Folder\OpenDOS\Disk Cache Driver" 1024 16384
> > Don't you think it's a bit wasteful to load all the drivers beforehand? A
> > utility to mount/unmount the new FS should be created - just like in Unix.
>
> Rather, one should have the choice. In Linux, I've got minix,
> ext2, MSDOS, UMSDOS, compiled INTO the kernel, and VFAT, ISO9660,
> and others compiled as loadable modules. This way I can choose
> what I think is important.
Hmm... I think you're right. But the default for the system dist should be:
one FS compiled into the kernel, all the others as loadable modules (again,
for *users* not programmers)
> > a) Yes, that's right. The only IFS documentation I know of is in the Ralph
> > Brown's Interrupt List, and it's IMHO insufficient to write an IFS driver.
> > It's easier to do that with Win95 as the VxD IFS interface is clearly
> > documented, but who wants (of us ;-)) to help improving an M$ product? ;-)
>
> Not me! A new interface is needed.
Vivat!
> C or ASM is fine. I personally would prefer that DOS code is
> written in ASM as much as possible though. Reason: SIZE!!!!! I
Right, ASM is required here. After all we don't have to think about porting
the system to other HW platforms, do we?
> have a 19k mouse driver. My friend's mouse driver is 9k. They
> are both dated from the same time. Why is mine 19k? Probably
> because it is written in C! (It is, I've examined it's memory
> wastage in memory with debug).
Where's the golden art of writing big apps in assembler...?
********************************************************
For when it comes down to it there's no use trying to
pretend. For when it comes down to there's no one really
left to blame - blame it on me, you can blame it on me
We're just sugar mice on the rain.
----
Visit http://ananke.amu.edu.pl/~grendel
- Raw text -