Ross Harvey wrote:

> Chris did more than just the original alpha bootblocks.
> 
> IIRC, you said that you don't read source-changes, so you may not be aware
> that quite recently Chris completely rototilled the alpha boot blocks and
> installboot. A big part of that project was specifically cramming the ffs
> and cd9660 name-lookup primary into 7.5K. The LFS work was even more recent.

That was one question I had about the alpha.  Is the area _before_ the
bootblock info and the end of sector 0 available for code?  On the pmax,
we really have a 8k-64 limit on code size, not 7.5k, but obviously this
makes sharing sector 0 almost unmanagable.

> Furthermore, he answered your question, in part, earlier in this thread.
> It's easy at the hand waving, concept stage, to declare a broad equivalence
> between root and subdirectory searching, but Chris noted here that one
> practical difference is the new requirement for a more functional close
> routine.

I saw this comment before and my first thought was "why?". libsa
currenly can open a maximum of SOPEN_MAX (4) files.  Just a plain
subdirectory scanner doesn't use open files, and my bit of pseudocode

	if open("/boot/${MACHINE}") fails
		if open("/boot") fails
			die

only uses two of these at worst case...

Simon.