Wed 2006-06-28 13:46, Maurice Kinal (1:261/38.9) wrote to andrew clarke:

 ac>> It's much easier to fix broken source code than to try to get 
 ac>> everyone to use newer file formats for everything.

 MK> Who is 'everyone'?  Seems to me there aren't enough people using
 MK> any ftn software nevermind the older abondonware (shareware) except
 MK> the OS/2 guys who seem to like emulating DOS-think.

Everyone using the existing formats.

 MK> If nothing is done about it I doubt the situation will improve with
 MK> time and will no doubt cause a further decline in the nodelist.

It won't matter.  The world has moved on.

 ac>> I don't have any experience with using GNU C on 64-bit CPUs but in 
 ac>> any case, from what I understand it's possible to tell the compiler 
 ac>> to use 32-bit ints.  Pointers will still be 64-bit, but that 
 ac>> shouldn't cause any problems.

 MK> Yes but 32-bit ints break ftn software which require 16-bit ints. 

All of the code should have been written to not care about sizeof(int) by
now.  Some of this work was done with Paul Edwards' port of MSGAPI, more
than 10 years ago, which the Husky stuff now uses.  Later, Tobias Ernst
worked on some of it to remove the requirement for a particular endianness,
so he could get Msged working under Mac OS X.  There's not really any
reason I can think of to stop any of it working where sizeof(int) == 8.

 MK> You'd have to use a bit shifting method on a pure 64-bit platform
 MK> to compensate for DOS-think or have two sets of libs available
 MK> which seems to be the current strategy.

It already does bitshifting.  I assume there are two libraries presumably
because you can't mix 32-bit & 64-bit libraries in Linux.

Anyway, I think you're jumping to a few too many conclusions here.  Gert
might have discovered a new bug.  Unfortunately he's not very good with bug
reports and I don't understand his English, so I can't help him.

 MK> Funny thing is I was using 64-bit systems well before writing code 
 MK> for 16-bit DOS and never had this problem until Fido.  I wonder 
 MK> how many other networks will be byting the dust given the lack of 
 MK> poor implimentation of code.

IPv4 uses 32-bit ints.  The endianness is fixed too.  Doesn't seem to be
causing the Internet any problems though.  :-)

-- mail{at}ozzmosis.com

--- timEd/FreeBSD 1.11.b2