On (07/20/07 22:27), Peter Memishian wrote:
>
> Not to me. The other thing that seems inconsistent here is that existing
> (and upcoming Clearview) show-* commands are structured so that each row
> represents an object of that type -- e.g., for show-link, each line is a
> link, for show-linkprop, each line is a link property, and so forth.
Current show-<link> commands output (1 X M) matrices with each row
corresponding to one link, whereas I am proposing an M X N matrix (M !=
1) for the link.
If we constrain ourselves to the 1 X M model, then we are back to
autoneg,cap autoneg,curr autoneg,adv speed,cap speed,curr ...
<link>
which is definitely not going to be human-readable.
With the current output formatting rules, the row information is a
constant for the table, whereas the number of columns is completely
dependant on link type, the number of properties associated with the
link, and someone's judgement of which properties to display.
Transposing the matrix (to M X 1) would at least allow us to not face
the spilling-over-80-columns mess, though, of course, the machine-
and human- parseable output are no longer easily derivable from each other
any more.
At some point, one starts to wonder if we are trying to get useful
information across, or trying to conform to some set of output
formatting rules, because the two objectives are starting to get at
odds with each other :-(
I personally feel that we should allow the machine-parseable output to
be just that, and not impose constraints on the human parseable version.
If we allow ourselves that, then one possibility is to have the
machine-parseable (-P) output display something like
(row1,column1)=(value) (row1,column2)=(value) ... (row1,columnN)=(value)
(row2,column1)=(value) (row2,column2)=(value) ... (row2,columnN)=(value)
:
(rowM,column1)=(value) (rowM,column2)=(value) ... (rowM,columnN)=(value)
For the 1 X M matrix case, the row is going to be repetitive information,
but then again, this is machine-parseable output, so it's not going to
be an eye-sore to anyone.
In this model, we could go with the tabular output I proposed in my
previous mail (which I'm not too fond of, either, but is at least
preferable to trying to dump all that information into a 1 X M matrix).
--Sowmini
