Ethereal-dev: Re: [Ethereal-dev] Updates and a new feature

Note: This archive is from the project's previous web site, ethereal.com. This list is no longer active.

From: Guy Harris <gharris@xxxxxxxxx>
Date: Fri, 07 Oct 2005 02:16:49 -0700
Greg Morris wrote:

I have a number of new additions to Ethereal that I want to submit. But I know that the list was wanting only bug fixes this week prior to the next release. Should I still submit my new additions or wait for another week or so?

I'll let Gerald give an official answer, but it appears that he's branched off 0.10.13; if so, it's probably OK to add new features to the main branch, as they won't go into 0.10.13.

New Feature:
Error Equivalency Tables:
Similar to SRT functionality, but to enable, each dissector needs to add an additional tap. (<protocol>-err) IE tcp-err. This feature parses the packet trace and via the tap interface counts the number of errors reported and provides a table of the composite data.

Should those errors (or, at least, the significant ones) also be marked using the "expert" feature Ulf Lamping added - or should *repeated* errors be so marked? I think some network analyzers' expert feature reports problems such as "too many open errors".

I have currently done the error table tap and modifications to both the NCP and TCP dissectors. One neat feature is the ability to highlight one of the errors listed in the error table, right click the mouse button, and then search the internet for the error string. (I currently default to Google for the internet search)

Presumably it uses the "browser_open_url()" routine declared in gtk/webbrowser.h, so that it'll open the URL correctly on:

	Windows;

	OS X;

UNIX+X11, if the preference for the browser "open a URL" command is set correctly (someday, if we support a GTK+GNOME version, that version could use the standard GNOME APIs to let you open up a browser with a given URL, and if we ever do a Qt+KDE version, it should also do that).

I have attached a few screen shots of this new feature.

It looks as if it provides, in some ways, a summary window for the "expert" mechanism (especially given that the TCP statistics aren't all for errors - window updates aren't really errors).

There's currently no summary mechanism for the expert feature; perhaps the expert feature should support both a "generic" string for items, such as "Connection establish request" for TCP, and a "specific" string for each item, with the latter being what's displayed in the "Expert Info" window, and the former used to compute the summary counts for the "error equivalence" window and to do the Web search.

We might be able to piggyback your new feature atop the "expert" mechanism. Does that make sense?