Back to home page

DOS ain't dead

Forum index page

Log in | Register

Search:  
Back to the forum
Board view  Mix view

"Black Box" Approaches. (Announce)

posted by bretjohn Homepage E-mail, Rio Rancho, NM, 22.06.2010, 18:26

My further opinion, FWIW.

> > The "wrapper" could have contained allocate/deallocate entries for cache
> > unit numbers. The "wrapper" could have contained the "AMIS/IISR" logic
> > that people who want "new style" drivers so urgently desire -- I do not.
> This I can understand as back when I wrote a number of TSR's (mostly
> internal home/work projects) I studied AMIS and personally felt that it was
> overkill which added complexity that I didn't need (at least that's what I
> felt at the time).
Is AMIS overkill? Perhaps, but it is the only viable option available that I know of. As I stated earlier, the only other alternative to AMIS I've ever seen documented in TesSeRact. While similar in concept to AMIS, TesSeRact had one insurmountable problem -- it continued to use INT 2Fh, which contains all of the Microsoft baggage that comes along with it. I know Wikipeida talks about "several other competing specifications", but I've personally never seen any of them. Maybe the only reason AMIS seems to have "won out" is because it is supported by Ralf Brown in the infamous RBIL. Whatever the history, AMIS seems to be the only viable choice today.

The reason programs use IISP and AMIS is not so much for themselves, but rather for other programs, and ultimately the user. AMIS/IISP allows other programs to insert their interrupt handler "before" yours if they want/need to, even if you were actually installed first in time. Likewise, it allows another program to be uninstalled even if you were installed after it was and are both using some of the same interrupts. In other words, the protocols allow drivers and TSR's to "play nice" with each other, which ultimately benefits the user.

It is true that AMIS requires a few hundred bytes of resident overhead, but I personally believe it is well worth the cost when compared to the benefit it provides the user.

I don't think it is ever a programmers right to decide that a particular programs functions or services are so "indispensable", or that is uses so little memory, that the user will never want to remove it. That ultimately and always should be the users decision, not the programmers.

> I also remember thinking that if TSR's were written correctly AMIS didn't
> matter.
The only way that would be true is if a particular program only used interrupts or hot-keys that no other program would/could ever possibly use, which is usually impossible to guarantee. I think the only way to guarantee something like that would be if the program didsn't intercept any interrupts at all. I don't know of a "useful" device driver or TSR that doesn't intercept any interrupts, though it's possible there is something out there.

Post reply
 

Complete thread:

Back to the forum
Board view  Mix view
22649 Postings in 2111 Threads, 402 registered users, 812 users online (0 registered, 812 guests)
DOS ain't dead | Admin contact
RSS Feed
powered by my little forum