John McKown wrote: >What I, personally, think would be nice is something which exists,
>somewhat, in z/OS UNIX files. That is "file tagging". In z/OS UNIX you can >"tag" a file's contents to specify that the text in the file is in a >particular CCSID (code page). What might be nice would be if every piece of >data could, optionally, be "tagged" as having a CCSID for the data inside >it. Of course, this only works when all the data inside is "textual" and >not "binary". Anyway, tag the data as being in a particular CCSID. Then >have a parameter on the DD statement which basically indicates the CCSID >that the program expects its data to be in. If the data doesn't have a >"tag", assume CP-037 for historical purposes. If the DD doesn't specify a >CCSID, assume CP-037 too. So in the default case, the CCSIDs would both be >CP-037 so no translation would be made. I am assuming that the access >method would do the translation, using Unicode System Services, and would >not do a conversion if the CCSIDs were identical. If the CCSIDs don't >match, then the access method would use the Unicode services to do the >translation on input & output. I am assuming that all members of a PDS >would be in the same CCSID. I guess this could be extended for individual >members in a PDSE version "2.1" or something. The tagging would be >automatic if a file is created using JCL or DYNALLOC by using the CCSID >specified in the DD or DYNALLOC. Again, defaulting to CP-037 for historic >purposes. >I'm not saying how this "tagging" would be implemented. I don't know if >there is any room in the VTOC entry for a CCSID. Or if it would require >something in the VVDS. And the problem remains for "mixed" data where the >records contain both "binary" and "textual" information. That's an intriguing thought. Of course adding metadata like that to z/OS data sets creates a whole new mess, with "Oh, that application/whatever doesn't handle the tagging", but once we got there, it sure would make life simpler! ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
