Tagged fields make alot of sense because a parsing program can easily
recognize the fields it needs and ignore the rest completely. The
syntax you suggest is probably best, especially since many files
already use it and it is similar to mail files. More important is
that whitespace be allowed before (for readability) and after the
value (for padding).
What needs to be standardized for the procfs is the basic common
syntax and the field names/types. Programs that read /proc data can
then have a general parser that separates a file into fields. Then it
can find the fields that it needs and convert the data into basic
types. The atomic types could be: decimal, hex, real, string (quoted
arbitrary values), and alphanumeric. We might want to come up some
documentation for programmers of the standard names and types for each
field in each file.
There also needs to be a syntax for tabular data. My suggestion is
have the first line identify the table name and fields and each
following line contain a record with each field separated by
whitespace. For example:
arp(ip_addr, hw_type, flags, hw_address, mask, device)
36.178.0.1 0x1 0x2 00:00:0C:49:40:9A * eth0
This just standardizes the existing tables, some with headers and some
without and names the fields.
The best way to implement this is write an interface so that most
programmers don't need to worry about formatting things:
print_field_alpha("cpu", "i386");
print_field_real("bogomips", bogomips);
- Ian
--
-- Ian Burrell == iburrell@leland.stanford.edu **
<URL:http://www-leland.stanford.edu/~iburrell/>
Simon's Law: Everything put together falls apart sooner or later.