[Twisted-Python] RE: In Defense of Taps

L. Daniel Burr ldanielburr at earthlink.net
Thu Feb 13 14:38:54 EST 2003


Just to throw another newbie perspective into the Twisted mix:

I like the idea of taps.  As Glyph has mentioned numerous times,
the existing tap concept is not as good as it could/should be.
Even so, I find taps useful for one simple reason: I can sell
Twisted to my boss because of taps and plugins.  If I tell him
that I want to use this great networking framework, but I am
going to roll my own custom code to actually start/stop/execute
the application I am building, he's going to shoot me down.

On the other hand, if I tell him I am going to build my nifty
application, using a nifty framework, and the end result will be
something that can setup on our dev, test, QA, and staging
environments just by deploying the code and typing
"mktap -foo -bar -baz", "twistd --file=mytap.tap", then I am
much further along in my quest to gain approval.

Taps and plugins are a Good Thing, for reasons other than just
technical.  I can certainly see why many developers would say
that taps are something they don't care about, because they just
want to create a reactor, hook up a factory and protocol, and go.
In my case, I want to create a plugin, and have somebody *else*
be able to hook it up and go.

Bottom line for me is that taps and plugins make Twisted a more
finished architecture.  Java servlets would not be so widespread
if all developers wrote their own code to start the container,
configure the context, etc.  Quite the opposite actually: They
drop their code in a webapp directory, edit their web.xml and
go.  Taps and plugins sort of do that for Twisted, which is all
to the good.

L. Daniel Burr




More information about the Twisted-Python mailing list