Article 3660 of alt.sys.pdp10: Path: nntp1.ba.best.com!inwap From: inwap@best.com (Joe Smith) Newsgroups: alt.sys.pdp10,alt.folklore.computers Subject: Re: 36-bit MIME types, PDP-10 FTP (was Re: File transfers was Re: Barb, DSKB is not in catalog!) Date: 26 Feb 2001 23:56:58 GMT Organization: Chez Inwap Lines: 54 Message-ID: <97eqga$175u$1@nntp1.ba.best.com> References: <3A85706B.B4A96DC3@nospam.nospam> <96k6vk$a33@nnrp3.farm.idt.net> <3A8DA401.D850A6BB@unisys.com> NNTP-Posting-Host: shell3.ba.best.com X-Trace: nntp1.ba.best.com 983231818 40126 206.184.139.134 (26 Feb 2001 23:56:58 GMT) X-Complaints-To: abuse@best.com NNTP-Posting-Date: 26 Feb 2001 23:56:58 GMT Xref: nntp1.ba.best.com alt.sys.pdp10:3660 alt.folklore.computers:16823 In article , Brian Inglis wrote: >On Fri, 16 Feb 2001 16:04:49 -0600, "David W. Schroth" > wrote: > >The DEC OSes used 5x7 packed ASCII on 36- and 18-bitters (a pain >on the latter, from personal experience in assembler). Nine bit >bytes seems like a cleaner approach. >Does TOPS-20/TWENEX support FTP and how does the binary mapping >(36 bit<->octet) work there? Any references? Yeah, I posted one back in December. To whit: |>I knew I'd read it somewhere... Still not sure how TOPS-20 turned two |>words into bytes (as with binary files on wsmr-simtel20). Perhaps it |>stored 9 bytes in 2 words? I did some experimenting with TOPS-20 to Unix transfers. For TEXT mode, use a 7-bit byte pointer to read or write the 36-bit file. The LSB on the 36-bit word is not sent, and is zero on receive. For IMAGE mode, use a 4-bit byte pointer and read two nybbles for every octet sent (or write two nybbles for every octet received). (The least significant 4 bits of the first word would be combined with the most significant 4 bits of the second word.) The end result is that two 36-bit words (72 bits) are encoded as nine octets (72 bits). For TENEX mode, use 8-bit byte pointer (which will skip over the least significant 4 bits of every word). Suitable for 8, 16, or 32-bit data. For PAGE mode, there was a separate negociation between the two TOPS-20 systems to skip pages in sparse files. If an 8-bit program (such as a CP/M binary) was stored on a TOPS-20 system (such as SIMTEL20) and downloaded to an 8-bit platform in IMAGE mode instead of TENEX mode, the resulting file would be mangled. But the data could be recovered. I remember seeing TENEX.COM, a program for CP/M that would read a mangled input file 9 bytes at at time and write 8 bytes at a time (after discarding the 2 unwanted nybbles). KERMIT transmits 36-bit words as four 7-bit characters plus one 8-bit character. (The high-order bit of the 5th octet matching the least significant bit of the 36-bit word.) A TOPS-10 .EXE file could be transmitted to a CP/M system and back again without losing any bits. -Joe -- See http://www.inwap.com/ for PDP-10 and "ReBoot" pages. -- See http://www.inwap.com/ for PDP-10 and "ReBoot" pages. Article 3645 of alt.sys.pdp10: Path: nntp1.ba.best.com!news1.best.com!news2.best.com!feed2.news.rcn.net!feed1.news.rcn.net!rcn!not-for-mail From: "Alan H. Martin" Newsgroups: alt.sys.pdp10 Subject: Re: Kermit (Was: Barb, DSKB is not in catalog!) Date: Sun, 25 Feb 2001 22:45:39 -0500 Lines: 51 Message-ID: <3A99D163.458F5AEE@MA.UltraNet.Com> References: <3A85706B.B4A96DC3@nospam.nospam> <963uq7$b9s$1@spies.com> <9643f0$ohe$1@newsmaster.cc.columbia.edu> <96e538$d0h$1@newsmaster.cc.columbia.edu> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Trace: UmFuZG9tSVaZK5WUcPRM8Pa3Bi1nSYBaCtKeO17H7ktVVZzfHdRSCBPUoj9Im3Z4 X-Complaints-To: abuse@rcn.com NNTP-Posting-Date: 26 Feb 2001 03:46:23 GMT X-Mailer: Mozilla 4.76 [en] (Win95; U) X-Accept-Language: en,en-US,en-GB,es Xref: nntp1.ba.best.com alt.sys.pdp10:3645 Frank da Cruz wrote: > > In article , > =?iso-8859-1?q?Bj=F8rn?= Hell Larsen wrote: ( ^^^^^^^^^^^^^^^^^^^^^^^^ This is probably ironic). > : I would rather have developed a Unix utility that was able to read > : and write files within an emulated T10/20/etc disk drive/file system. > : > : If would be more work, of course, since Kermit already is here and > : working fine, and since serial line support needs to be implemented > : anyway, but it sure would be useful. > : > You can't just map the TOPS-20 file system to UNIX. If you thought > PC-based NFS was a nightmare with its LF/CRLF confusion, imagine > NFS-mounting a TOPS-20 disk on UNIX. The text files have five > 7-bit bytes packed in a 36-bit word with one bit left over; the > binary files all 36 bits. Any such scheme would have to not simply > "map" the files, it would have to totally rewrite them, which is what > Kermit already does. Plus any kind of automatic scheme would have to > be done by guessing which kind of file each one is, which would only > work part of the time. So you guess. The first release of the VMS NFS server I used (UCX) would serve up stream-LF files to Alpha/UNIX just fine. However, it didn't translate variable-length RMS files. (Or, it wasn't configured to do that). So people trafficked in tiny Unix RMS conversion filters, and I hooked GNU Emacs find-file to detect and convert likely buffers full of RMS files. One day someone changed something on the VMS system, and Emacs stopped prompting me to convert buffers on files that were still in RMS format. Obviously a heuristic was kicking in. VMS files have way too much file and record attribute info, but the TOPS-20 FDB (and eventually even the TOPS-10 RIB) have those byte sizes. Were many binary files marked as 7-bit - some cusp might drop bits on the floor. Some text files were marked 36-bit (BASIC-10 files with LSNs; the DECUS tape files originally provided to Trailing Edge; ...). Time for a heuristic. At least LSNs are easy to detect; all you need to look for is in the first 7 characters of the file, and it's no more dangerous than Unix leaning on magic numbers. Did Cobol eventually set the byte size to 6 for DISPLAY-6 output? /AHM -- Alan Howard Martin AMartin@MA.UltraNet.Com