On 19 Apr 2014 18:51:09 -0700, in bit.listserv.ibm-main you wrote: >IMO, programming language and file organizations are >complete different things and should be discussed seperately. > >Furthermore, you have direct access on all platforms, including >Windows and Unix; this is not a matter of programming language at all. >And: you have keyed access, because it is easy to implement keyed access >by software on top of the hardware supported direct access; you don't >need hardware support for keyed access ... this was one of the design >errors in history.
While this is true in the sense that there are both indexed and direct files in the Unix and Windows worlds, so far as I know there is no common format and access routines provided with those operating systems such as BDAM, ISAM (remember that?), VSAM RRDS and VSAM KSDS are provided by VSE and MVS. As I understand it most if not all of the BUNCH provided similar facilities. Also my understanding is that in the Windows environment separate facilities such as Btrieve and COBOL vendor provided routines were needed with no general readability except as provided by those packages. This has meant in at least one UNIX shop that I worked at, all mainframe VSAM KSDS files had been converted to Oracle tables as part of the migration from the MVS system to HP-UX. It is the requirement for each shop to construct or purchase separately their own routines for this and various packages may use different routines with resulting incompatible file formats. As klutzy as the IBM utilities are they are still more than what is available on my Windows 7 computer. Clark Morris > >I constructed a library of my own which allows to do keyed access from >C programs (different flavors of keys and different flavors of key >access logic), >which is in use at a very large customer of mine since ca. 15 years now and >which outperforms some other libraries and databases doing the same >(on Windows, Unix and z/OS, by the way). On z/OS it is built upon normal >data sets with RECFM=FBS, constant record length, and the physical reads >are done using fseek and fread (functions of the ANSI-C runtime library). > >This is no rocket science; the methods how to do this are known since at >least 1972 >(not in C then, of course, but, as I said already, it is no matter of >language). > >If you want to know more about that, contact me offline. > >Just another story: I recently had to document the internal structure of >keyed files >implemented by an old compiler for Windows called RM/COBOL; the compiler >dated from >the mid 1980s and is today marketed by MicroFocus. I was able to understand >the internal structure in the end (there were some informations on a web >site >in Russia) and to write a C program to read the RM/COBOL index files >and convert them to "normal" flat files. That is: there were already >implementations >for keyed files on Windows very early ... when there was no Windows yet ... >only PC-DOS. > >Kind regards > >Bernd > > > >Am 20.04.2014 02:44, schrieb Clark Morris: >> Frankly I agree with Pete Dashwood who posts on comp.lang.cobol when >> he says COBOL is dying, though not with quite the same analysis of the >> situation that he provides. This is especially true in the Windows and >> Unix areas where file organizations that were basic for both IBM and >> the BUNCH are not provided (direct and keyed). On the z series, it is >> items like this that lend credence to this belief. Clark Morris > >---------------------------------------------------------------------- >For IBM-MAIN subscribe / signoff / archive access instructions, >send email to [email protected] with the message: INFO IBM-MAIN ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
