Sean McGrath, CTO, Propylon

Sean McGrath's Weblog.

Friday, August 18, 2006
    Prime numbers are randomly distributed
Not. I might have blogged this before some years ago...I recall seeing it before.

I stumbled upon it today using http://www.stumbleupon.com/ - a service to which I am becoming rapidly addicted.


posted by Sean 9:12 AM
[Link]
. . .
Wednesday, August 16, 2006
    Ielnrcdibe?
Reading Test.


posted by Sean 5:22 AM
[Link]
. . .
Tuesday, August 15, 2006
    The (non) wisdom of crowds
    "So, rather than competition driving true technical innovation to greater and greater heights, market forces can distort innovation necessitating a slavish attention to the simplistic metrics the market has adopted. " -- The (non) wisdom of crowds


posted by Sean 2:39 AM
[Link]
. . .
Monday, August 14, 2006
    Did you send me an e-mail in the last two weeks?
If so, I might not have got it. I'm gradually assembling evidence that something bad happened to e-mail heading my way during first two weeks in August.

Please resend any e-mail sent during that period.

Ta.


posted by Sean 9:38 AM
[Link]
. . .
    The XML Summer School 2006
The XML Summer School was very interesting and enjoyable again this year. Highly recommended.

This year, I talked on the subject of Web Services and SOA.

At some point during my talk, Paul Downey took a photo.


posted by Sean 2:17 AM
[Link]
. . .
Sunday, August 13, 2006
    JINI, Asynch Message Queuing and all that temporal decoupling stuff
An old hobby horse of mine this temporal decoupling thing. I cannot help beating on that particular drum as I see it as a really important concept that deserves to be better known.

Take a squint at OSCS for example. Using Amazon's S3 to splice a JINI-style asynchronous IPC layer into your application with very little heartburn compared to actually using, say, JINI locally. (JINI is powerful stuff but the last time I set it up, it cost me a week. Things may have improved since then.) This sort of infrastructure is a pain to set up and costly to keep running. If the cloud can do it for you...

Then there is the idea of using this layer to build an async message queue abstraction and using the message queues to orchestrate scalable web services where the processing can be done on any old machine (or machines) while the message queue layer, hosted in the cloud, does all the scalability/reliability etc. etc. Again, if the cloud can do it for you...

There are some patterns here which will grow in importance and popularity over the next few years. Are we seeing the virtualisation of the middleware layer? The outsourcing of the "ilities" of web apps?

Obviously, for some application areas this sort of out-sourcing of the middleware layer will not be a runner but for other application area, I think the whole area has significant legs.

To be provocative : A trendy new mashup without an asynch message queue handling all that nasty non-functional aspects of application design (like scalabilty, reliability, availability etc.), is either (a) only half a mashup which will generate blue smoke if (shock horror!) it becomes a success or (b) a full mashup that is way too expensive to operate and run to be around for the long run, unless its owners get subsumed into some mothership or other.


posted by Sean 12:52 PM
[Link]
. . .

. . .
Weblog Commenting by HaloScan.com