So Tim Bray was giving a larger architecture argument why AJAX was good, as opposed to the usual user experience goals: Better distribution or resources - you can do simpler faster requests on the server if the server doesn't have to render the interface, but can leave that to the client. To help in that process Brian IngersonIngy has moved even more processing onto the client by adding an implementation of the Template Toolkit common in perl web applications.
Of course it really gets crazy when people start taking the various "Javascript as VM" seriously. The perl 6 via Haskell effort Pugs has a javascript backend so you can run your entire application client side.
It's interesting that all of this can actually sensibly be combined with web linkyness and RESTful principles to make a truly application dissociated web - like the one we've been dreaming of for a couple of decades.
If you are running Windows on your PC you will need to reinstall it once in a while if you want it to perform decently. After a reinstall you will probably spend at least a day downloading files and clicking through installers. This nightmare is over now. Install windows-get and you will be drinkin' coffee on the balcony while your computer downloads and installs all the software you need. Just as if you were running apt on a Debian system.
UPDATE: Get a list of applications that install silently using win-get here.
To install an application silently (without interaction) from this list do
win-get sinstall appname
(e.g. "win-get sinstall skype" to install skype)
- If no silent install option is available win-get will run the GUI installer.
We really, really need some more physical interfaces for computers. Better haptics. I'm as bowled over by the Minority Report-esque interface demo that's making the blog rounds, but why can't I have that interface now, Mom? Physical interfaces done right would mean more than just improving the sensory interfaces we have. Better physicality of the interface also require a purification of the ideas in the interface so objects in the interface interact in a more physical way (this is quite apparent from the demo btw). The problem is we're not discovering these models because we don't have the physical interfaces that make the flaws of present interfaces obvious. So what can we work on today? One possibility is the admirable and quite easy to use Processing. And while processing doesn't have a funky enormous touch screen, it is able to use your webcam a la eyetoy, to play with fire - or snow.
The upstart Krugle will offer a search engine for source code. If this gets done right with suitable metadata this could totally change the landscape of software development. Of course it could also be done in a bland manner and make no difference at all. So what would be right? Well - total reflection in a manner suitable to each indexed language is a must. Documentation in source code should be digested if it comes in a standard form as well. And we should have strongly typed metadata wherever pertinent.
Arguably all of the SOAP interface discovery protocols out there have failed to bring a coherent market for web services. At the idea level at least Krugle could be active here as well.
(and yes, this is yet another one of those millions of ideas many of us have lieing around on lists at home. I'll cross this off mine)
I would love to have a simple fast loading free text outliner. The outliner in MS Word is the one I would like beaten with free/open/simple software. Tree outliners suck. A lot. I would also love a simple fast loading table/csv file editor that beats Excel. It would need to load CSV files into cells with all the manipulexity of Excel and it would need to save them without the idiotic "upgrade to our file format" save cycle of both MS Office applications and Open Office applications. If the latter app also was good at opening and manipulating SQLite databases (as tabbed sheets) that wouldn't hurt either. Any thoughts?
Please don't say "OpenOffice" as an answer for this. I have footprint issues with these, AFAIK there's no outliner in Open Office tect and they do run you through the annoying "would you like to 'upgrade' to our format"-cycle. Also, if you say "Vim" or "Emacs" you need to qualify your statemet with proper scripts/modes and if you say "Get a Mac" then you need to...say it somewhere else. I'm waiting for the first update of the Intel processeor Macbook Pro - and VMWare for the Mac - and then I'll switch.
One of the annoyances I've had with Catalyst (apart from a hellish install on my old Redhat server) is that the developers didn't seem to get the Rails philosophy even though they were building a rails clone. Fortunately it seems a lot of the problems have been covered by other hackers (god bless CPAN) through the introduction of Catalyst::Enzume one stop shopping for Catalyst::Crud. CRUD for Class::DBI as well as sensible template defaulting helps tremendously when speed is of the essence. And speed is always of the essence.
The Io language is small, cross platform, highly readable, carefully designed and dynamic. In addition, the perspective of the developers of the language seems to be Just Right. To be investigated.