Argus Printf Statement

[Carter Bullard]

Hey Russell,
   Yes, tabs seem like they have more utility as
an alternate default delimiter.  Getting the data
into Excel, for example, is a lot easier if you've
got tab or comma delimited fields.

   We'll want to do this in addition to the argprintf()
type approach, which will give us a lot of flexibility
to specify whatever delimiter you want.  The task of
defining the syntax is the hard part, and so we'll need
to start that soon. Anyone want to take a first stab?

   It seems that an Argus-1.0 question should be asked
again, and that is what should the default ra() ASCII
string be?  We all thought hard to get a default 80
character output string that looked OK and had what
was considered useful information.  And we introduced
the -W option for when there wasn't an 80 character
limit for the output device.

   This is still very important!  Do we want to
modify the existing default output strings?

Carter

-----Original Message-----
From: owner-argus at lists.andrew.cmu.edu
[mailto:owner-argus at lists.andrew.cmu.edu]On Behalf Of Russell Fulton
Sent: Monday, July 17, 2000 1:36 AM
To: argus at lists.andrew.cmu.edu
Subject: Re: Argus Printf Statement


FIrstly sorry for the silence when this first came up -- I have been 
off the net for a few days....


On Wed, 12 Jul 2000 08:03:15 -0400 Carter Bullard <carter at qosient.com> 
wrote:

> 
> This I believe would be a huge thing to do, and I would like to
> get some opinions on how this could work.  Now I don't have any
> Perl experience, so all my examples will be C oriented.
> 
> I can see providing an argprintf() function that mimics sprintf():
> 
>    argprintf((char *)buf, (char *)formatstr, (ArgusStruct *) arg)
> 
> and the formatstr can have a syntax very much like printf() and
> strftime().  A first thought, we could come up with a syntax
> that allows us to extend the normal printf() and strftime()
> formats with Argus data identifer tags.  This would allow a
> preprocessor to be able to construct real sprintf() and
> strftime() calls based on our syntax.
> 
> We've got to be able to specify source vs. destination for metrics
> and flow identifiers, so a %s.X and a %d.Y type of qualifier may
> be all that is needed.  For time we've got start and stop time
> values and their formats to consider.

Yep, that basically what I had in mind.  In particular I wanted to be 
able to supply the format string to whatever replaces ra...

That said, I could live with the simple delimited lists output that 
others have suggested.  One point though is how do we get at data which 
isnt in the default displays? eg. time to live

One motivation for specifying the formatting is that formatting seems 
to be a major part of the cpu overhead of ra.  (i.e. ra spends a lot of 
time formatting output records)  So I thought that it might spead it up 
by only getting the data formatted that I wanted.  This isnt a very big 
deal though and may not be worth the effort.

As for what delimiter to use I prefer tabs since they won't occur in 
any legitimate data.  '-' are used in the current default format for 
the reset/direction symbols.  That said I really don't care so long as 
we can be sure it won't someday pop up in a data field.

Cheers, Russell