tornado & websockets

April 26, 2014

The core of my POC prototype is a server engine, which doesn’t really make for a great demo. Generally, people grasp concepts quicker if they can see a tangible realization. So I needed a realistic way to show live ticking data getting cranked out by the server. A browser GUI seemed a natural candidate. And being a Pythonista I wanted to do the server coding in Python. Until recently getting live ticking data pushed up to a browser was a big deal, requiring sophisticated server products like the Caplin Liberator, and rich GUI toolkits like Caplin Trader. Fortunately, it’s now possible to hack some demoware in the form of a live, ticking webpage using some really simple jQuery & websockets in the browser, and tornado on the server side. JavaScript and browser GUIs are not my forte, so I won’t comment any further, except to note how much easier it seems than five years ago. On the server side, though, I do have more experience. Back in 2000 I was doing server side web dev in Python using Zope. Zope is a very powerful system, featuring a built in Object DB and an inheritance by instance rather than class mechanism called acquisition. Consequently it has a rather steep learning curve. In recent years Plone has had some traction as a CMS built on top of Zope. In 2001/2 I discovered Twisted Matrix, a general networking toolkit you can use to build any IP based networking functionality. Again there’s a steep learning curve, but it’s much lighter than Zope, and is now very mature. I will be using Twisted to build a general socket server capability for my core product: I’ve got C++ and Python APIs, but I’ll need a socket server for Java support. But what I needed for my demo purpose was real time server push to the browser. And tornado proved to be a good choice. Simple, lightweight, lots of worked examples and focused entirely on websockets. It didn’t take long to get ticking data into a webpage. Recommended!

Advertisements

Nodally Xenograte

April 10, 2014

Well, xenograte from nodally sounds pretty cool: loosely coupled software components in the cloud. Brad Cox’s dream of snap together building blocks, finally realised. Yahoo Pipes, anyone? I’ve even hacked around similarly motivated code myself, but never got so far of course. The problem with these new paradigms is that they ask you to throw away all your old software assets so you can rebuild them again in the new framework. A bit like media companies asking us to buy the same content over and over on different formats: LPs, tapes, CDs, audio DVDs, downloads, pono, VHS, DVD, bluray….  Why can’t someone find a way to breath fresh life into existing assets without reengineering them. Why not, indeed?