Where should the type information be 164

On Thu, 31 Mar 2005 07:48:19 GMT in alt.folklore.computers, "Stephen

IIRC 3380 and 3390 had sector cells related to RPS which made space calculations bumpy (?14 byte cells + fixed overhead on 3380, ?232 on 3390), but 3390 was virtual CKD implemented on FBA devices, whereas 3310-3370 were 512 byte sector FBA devices, usable only on VM, maybe also DOS-VSE.

IBM CKD devices effectively reformatted each block as it was written, so each block contained an integral number of records, and the wasted space on each track was inaccessible to anything on the CPU.

Mainframe OSes had access methods and libraries, but not aware of minis ever having anything like that until RMS-11K (VSAM like) was introduced on PDP-11s; language runtimes supported text line and binary streams, but random I-O was only at sector level, so the programmer had to do the work to randomly access logical records spanning sectors, choosing between keeping two flip-flop sector buffers per file in memory for update speed, or rereading the previous sector on an update to reduce the buffer space used. PDP-11 addressing limited access to about 3 RMS-11K indexed files-program; however send-receive IPC could be used to do file access in server programs separate from the application, rather than link home-grown or bloated commercial libraries into the application.

ISTM that was an HLL term for CKD block I-O direct into user program buffers, which relied on integral records-block, and has nothing to do with any fixed sector drive technology, or async I-O, which is *used* by every *OS* other than single task PC "OS"es, but not always available to user programs except in mainframe systems.

-- Thanks. Take care, Brian Inglis Calgary, Alberta, Canada

fake address use address above to reply