home *** CD-ROM | disk | FTP | other *** search
- TCP-Group Digest Sun, 2 Sep 90 Volume 90 : Issue 127
-
- Today's Topics:
- Clarkson WD driver
- SMTP implementation ideas
- Turbo Assembler/Turbo C question
-
- Send Replies or notes for publication to: <TCP-Group@UCSD.Edu>
- Send requests of an administrative nature (addition to, deletion from the
- distribution list, et al) to: <ListServ@UCSD.Edu>
-
- Archives of past issues of the TCP-Group Digest are available
- (by FTP only) from UCSD.Edu in directory "mailarchives".
- ----------------------------------------------------------------------
-
- Date: Sat, 1 Sep 90 13:16:03 PDT
- From: Pete Carah <ames!elroy!grian!puffin!pete>
- Subject: Clarkson WD driver
- To: tcp-group@ucsd.edu
-
- I'm trying to get nos 731 running using the Clarkson WD packet driver.
- This is on a 286 with the board on irq 2 and all other defaults.
-
- The Clarkson driver is the latest one just fetched from sun.soe.clarkson.edu
- yesterday.
-
- The driver correctly redirects 2 to 9, so it spots the AT bus.
-
- PKTSTAT appears to show the right number of packets.
-
- My difficulty is that the receive upcall seems to not get called unless
- I send a packet, and then only sometimes. (The host has rwhod running so
- lots of packets get generated. My ethernet trace program works fine on
- the same AT with the same card, so I'm pretty sure that the card works right.
- It uses my own interrupt routine and not the "packet" driver's.)
-
- The only peculiarity is that I've compiled the nos using tc++ rather than tc,
- with some rearrangement of includes. Has anyone else used this combination
- of driver and compiler?
-
- I've tried later versions of nos and they have bugs in the socket code, so
- I'm waiting...
-
- -- Pete
- (pete@puffin.uucp) ...!{elroy.jpl.nasa.gov!grian | hacgate}!puffin!pete
-
- ------------------------------
-
- Date: 1 Sep 90 11:44 -0700
- From: Doug Collinge VE7GNU <djc@samisen@uvicctr.uvic.ca>
- Subject: SMTP implementation ideas
- To: tcp-group@ucsd.edu
-
- >X-Mailer: Mush 6.5.6 (PC R6.3 22-Sep-89)
-
- > From: Brian Lloyd <uvicctr!robin.telebit.COM!brian>
- > I have been using an SMTP client recently that has a very nice
- > characteristic. ...
-
- This is fine. But the nice thing about the "backing-off forever" method
- is that it has no user-detunable parameters.
-
- 73 doug [NO USER SERVICEABLE PARTS INSIDE - WARRANTY VOID IF SEAL IS BROKEN]
-
-
- --
- /\/\/\/\/\/
- Doug Collinge, first try: collinge@uvicctr.uvic.ca
- then try: samisen!djc@uvicctr.uvic.ca
- VE7GNU VE7GNU@VE7GNU.#VIC.BC.CAN.NA
- Victoria, BC, Canada
-
- ------------------------------
-
- Date: Sun, 2 Sep 90 05:12:05 UTC
- From: karn@ka9q.bellcore.com (Phil Karn)
- Subject: Turbo Assembler/Turbo C question
- To: shep@allspice.lcs.mit.edu, tcp-group@ucsd.edu
-
- Tim,
-
- You're quite right -- the arguments don't match, and they should. But
- things still work because there are no byte-wide stack operations on the
- 8088; calling a function with a character argument still pushes two bytes on
- the stack.
-
- The main reason I have recently gone to using ints (or shorts) in place of
- some of the chars in NOS is to avoid the $#@!! sign-extension problems that
- can occur since Turbo-C chars are signed by default. This problem has become
- much worse since ANSI-C came out, since type promotion is no longer
- "unsigned preserving" as it was in K&R C.
-
- I'll probably go through the code and try to clean up little nits like
- this. Ordinarily I find them through the type-checking that the compiler
- provides, but there is unfortunately no type-checking against assembler
- subroutines.
-
- Tangential flame on:
-
- In my opinion, the unspecified signed-ness of chars is probably the most
- serious design mistake in the C language. Other mistakes, in my opinion,
- include ANSI's changes in type promotion I just mentioned, the unexpectedly
- low precedences of the & and | operators, the default signed nature of
- integers, and the use of an undeclared function causing it to be declared
- 'int' by default. Chars and ints should both be unsigned by default,
- the & and | operators should have the same precedence as the binary arithmetic
- operators, and all functions should have to be declared before being called.
-
- I understand that there were good reasons at the time for some of these
- design choices. Chars were left unspecified because they were inherently
- signed on the PDP-11 (where C spent most of its childhood) and unsigned
- elsewhere. And the & and | operator precedences are as low as they are
- because there were no && and || operators originally in C and & and | were
- used instead. I understand that there were as many as 30 programmers already
- using C when the && and || operators were introduced, so they couldn't
- possibly make such a drastic change to such a widely used language...
-
- Phil
-
- ------------------------------
-
- End of TCP-Group Digest
- ******************************