Re: Nmap-dev-only release: Nmap 3.10ALPHA1

[Fyodor <fyodor () insecure org>]

On Wed, Aug 28, 2002 at 02:18:11PM -0500, William McVey wrote:
> On Wed, 2002-08-28 at 03:18, Fyodor wrote:
>
> Am I the only person who hates C++ and is sad to see the poC (plain ole
> C) version of nmap die?

The death of poC Nmap has ben greatly exaggerated.  Sure it compiles
with g++ instead of gcc, but at least 90% of the code is identical.  I
am not an OO paradigm fanatic planning a wholesale rewrite of the Nmap
codebase to include multiple inheritance, polymorphism, etc.  I have
just added the basics.  For example, there are new classes to
encapsulate data with related functions.  Constructors and destructors
have replaced ad-hoc calls such as options_init(),
hostgroup_state_init(), and hostgroup_state_destroy().  Syntactic
sugar like line comments (//) has been added, and accessor functions
have replaced some of the most bug-prone member variables to enforce
critical invariants.

The goal is to make the Nmap codebase more scalable, maintainable, and
robust without sacrificing performance.  Improving code organization
and readability will make performance-improvements to the Nmap
scanning algorithms much easier.  These algorithms can be hundreds of
times more important than optimization details like handcrafted
assembly packet checksum routines or reducing C++ object instantiation
overhead.

Give it a try.  If nothing else, you may appreciate the C++-style
comments :).

> It seems odd to me that a project that had
> conniptions on bringing in external libraries (libnet, libxml/libsax)
> due to portability and bloat issues would go down the C++ route...

I can't believe I'm being accused of bloat by one of the biggest Nmap
XML output proponents :).  But seriously, I don't see this bloat you
are referring to.  Nmap 3.10ALPHA2 includes IPv6 as well as the C++
conversion, and yet the source code size, executable size, and runtime
performance are very similar:

Package size:

-rw-------  1 fyodor fyodor 922293 Aug 20 18:19 nmap-3.00.tgz
-rw-------  1 fyodor fyodor 945497 Aug 31 18:45 nmap-3.10ALPHA2.tgz

Stripped Linux binary size:

-rwx------  1 fyodor fyodor 288560 Aug 31 20:47 nmap-3.00/nmap*
-rwx------  1 fyodor fyodor 306040 Aug 31 20:47 nmap-3.10ALPHA2/nmap*

Runtime performance on the same machine:

Nmap 3.00 64K localhost TCP scan: 3.22 seconds (average of 3)
Nmap 3.10ALPHA2 (same scan): 3.21 seconds (average of 3)

My tests have not shown any performance degradation from the C++
switchover.  If someone can demonstrate meaningful performance
problems with 3.10ALPHA2, I'll certainly investigate.

Cheers,
Fyodor

---------------------------------------------------------------------
For help using this (nmap-dev) mailing list, send a blank email to 
nmap-dev-help () insecure org . List run by ezmlm-idx (www.ezmlm.org).