Re: Question about UMBs

tom ehlert wrote:

>>The speculation was about finding a free block of memory in an area used 
>>for ROM shadowing. The resident code should be easily small enough even 
>>if there's only a single 512-byte page left in a 16K block. And if 
>>there's no block, I'll just take a small part out of the normal UMBs.
>>
>
>of course, you could move your code to c810:0, but don't forget
>to unlink from XMS chain when done.
>
>given that a straight UMB handler can be done in ~162(dez) byte, it can
>save you
>many headaches to remain resident in conventional memory.
>
I've already decided that the driver will just stay resident. 
Considering how little memory it would take to keep the UMB handler 
resident, the headaches just wouldn't be worth it. The loader will try 
to find a free block in shadow RAM, if there is any shadow RAM being 
used, and failing that it'll just take a small piece from one of the UMB 
blocks.