>From rlk@think.COM (Robert Krawitz) Mon Nov 30 10:56:46 1987 Let's see if I remember my BNF for babyl files; this corresponds to version 5: File :=

* ; Some say there must be at least one message. Header := Babyl Options:\n * |^_ Header-option := ; See note [5] : * header-token := [^\000-\017:\177-\377]* ; Not these characters [tab is OK] header-value := ditto, if a list, each element separated by a comma and a space. message := \^L\n [01], ; See note [1] below ( ,)* ; Note space before and comma after token , ( ,)* ; ditto, see note [4] below \n

* ; See note [1] and [2] below *** EOOH ***\n

* ; See note [2] below \n \^_ attribute := unseen | last | ; Not all programs implement this. It ; generally only gets used internally, and ; isn't written out to a file. >last | ; Babyl uses this for a deleted message at the ; end. It shouldn't be written out to a file. deleted | recent | ; Not all programs implement this. It refers ; to a message in the last batch of new mail; ; thus it probably shouldn't be written out to ; a file during a normal save although it ; makes sense to write it out in an emergency save. filed | answered | forwarded | redistributed | badheader | ; Not all programs implement this filed ; Not all programs implement this label := [^\000-\020,\177-\377]* ; No control chars, ; whitespace, commas, rubout, or high bit set header := [^\000-\020:\177-\377]*:

* header-line := [ \t][^\n]*\n ; Continuation lines must be indented body := (.*\n)* ; See note [3] below [1] A zero means that the headers have not been cleaned up, reprocessed, toggled, or whatever. In this case there should be no headers before the EOOH line. A one means that the headers have been reprocessed. In this case, the original headers will typically be before the EOOH line and the reformatted or whatever subset of headers that the user should see will be after it. Note that in this case it's permissible to garbage collect all headers before the EOOH line. No one's defined what it means to garbage collect SOME of the headers before this line, or what that means. [2] It's apparently permissible to add headers of the program's own choosing before the EOOH line. Or at least, Rmail does so (it caches a summary line) and nothing seems to object. There's no particular guarantee that something else won't step all over it, though. Headers after the EOOH line can be reformatted as the program wishes (e. g. indent the header lines to the same distance, canonicalize machine names) for display to the user. It's generally best for programs that read a babyl file to look at the headers before the EOOH line if they exist, since these should be untouched by the user. Remember, the user can edit anything after the EOOH line. [3] A \^_ at the beginning of a line should be quoted somehow. The normal way seems to be to decompose it into 2 characters: a ^ and a _. Strictly speaking, it doesn't always have to be, since the following text would have to be parsable as a message, but some programs don't try to use that much intelligence. Oh well. [4] Labels, or keywords as they are often called, are generally defined by the user, although it's not entirely impermissible for a program to use these for its own purpose (e. g. a keyword named RemindMe might be used to automatically find important messages). Some people also want these used to cache other state implemented by certain programs; this use is undefined. Note that all keywords used should be inserted in a header-option named Keywords:. Can a keyword have the same name as an attribute? Who knows? It's probably not a good idea, since some programs use the concept of = + . Sigh. [5] Some tokens are standardized in meaning. Common tokens are Mail inboxes, babyl file version number, which is currently 5, labels used in messages, window format for Zmail, anything else you want to be associated with a file. Be warned that labels should be a complete list of all user-defined keywords used in the file, so if you add a new label to a message, you should add it to this list. You should also have a Babyl version: 5 file attribute (look in a babyl file for details). Anyone know if there actually is a "formal" standard? This was done quickly from memory and a Zmail manual, but there are at least three programs around that use Babyl files (zmail, babyl, and emacs/rmail) and someone at SIPB was going to write a command-based mail reader similar to Unix Mail but operating on babyl files, and someone (of course not me :-)) should probably write xbabyl :-) References: ITS/Tops-20 INFO file on babyl (who wrote it? ECC? GZ?) Zmail manual (the MIT version was written by RMS; ECC wrote the section on Babyl file format) --