Yeah, because nobody wants to upload and store their photos on the web--so they certainly don't want to be able to edit them once stored there.

You're just missing the point by a mile. Online photo editing is a great feature for people who have already committed their entire photo collection to flickr (or one of the other several dozen competitors). There will come a time when photos are mostly taken by cell phone cameras, uploaded on the spot, and never dealt with except in an online form. An online editor is the only sane solution.

But, I agree that porting GIMP to forward its UI out to a browser is madness. Writing a GTK to JavaScript layer is probably significantly harder than writing an image editor from scratch in pure JavaScript or Flash (at least the 20% solution that'll satisfy 80% of users).

Wow, I am surprised I did not get downvoted into oblivion for my previous post. However, I am convinced that you're missing a point by a mile.

Nobody has committed their photo collections to flickr, because it is insane: people generate gigabytes of images with their cameras and upload only a tiny fraction of what they have, most of the time downsampled to a manageable size. Moreover, the sizes of image files (JPEGs) keep growing, while uploading speeds have stuck at 20-60kb/sec for 95% users.

This pattern is true not only for images, it's true for everything: the speed at which people generate data is only accelerating - from simple text documents we've moved to music, images, video and god knows what else. Sticking everything into the "cloud" will not work. Even today an average PC has tens of gigabytes of data. Soon that number will reach terabytes. Think about X-megapixel cameras, recorded TV shows, HD-movies, etc. The size of the backup drives will keep growing, while the speeds of internet connections will not.

Internet is a distributed computing platform with very powerful computing nodes (your computers). The opportunities of exploiting this are fascinating. SETI is a great example of using Internet right. Skype and torrents are another ones.

What these cloud companies are doing, however, is plain stupid in comparison: they're sucking in the computational complexity from "powerful nodes" only to re-distribute the load on their end onto their own nodes without any hope to keep up. What for?

I know that Microsoft is out of fashion today, but Ray Ozzie is absolutely right when he is talking about maintaining a reasonable (sane!) balance between client machines and "clouds". This point of view is not terribly popular among some YC readers, who are too fascinated and tempted by a slight chance of getting rich by re-implementing 2% of some perfectly fine desktop software in 2 months with a handful of python scripts running on cheap Amazon servers.

> "AJAX" is just one stupid function call

XMLHttpRequest() is one function call. AJAX in this context implies a rather serious JavaScript application. I note you didn't respond to the FreeNX alternate idea. Apparently there's already MozNX, a Mozilla plugin, so that would actually take care of the display (at the cost of another plugin).

I have an even better idea, though: A bootable Linux flash drive that connects to FreeNX running on your own EC2 instance with a deniably-encrypted filesystem (TrueCrypt). You can pop it in any borrowed computer and it just acts as a terminal, with your software running on EC2 and displaying remotely on your system.

Then you can have access to your apps and data anywhere and don't have to worry about spyware or keyloggers on unknown systems (except hardware keyloggers) and only pay for the hours you actually use the system.

(If you don't trust EC2 with personal data -- since you never know who else is going to be on your same Xen host -- then use whatever co-located machine you set up yourself, also with TrueCrypt; although without the machine in your physical control at all times, it could be compromised regardless -- not to mention trojans planted in your binary distro of choice.)

Again: Why? Why do programmers think that "from anywhere" matters so much? Sure "from anywhere" has some value, but it comes at enormous expense: unbelievable reduction in functionality. Please take a look at my previous post - I think that by pushing everything onto datacenters we're moving backwards, not forward.

Whenever you're designing an application, a true problem solver, you have the entire internet at your disposal. Your servers, and computers of your customers - all that combined RAM, CPU power and hard drives - they are all yours. USE IT and innovate. As opposed to taking something that already exists, adding "from everywhere!!!!" feature, slapping "wait... loading..." animation onto, and chopping 90% of the remaining features off.

Good example would be crap by Zoho. The proper way of doing it would be taking MS Office or Open Office and adding powerful collaboration features into them - the possibilities are enormous: you can have a rich UI, tons of processing power AND simultaneous document access, collaboration, sharing, backup, etc.

Instead we have a bunch of cowboy javascript coders reinventing a barely working wheel (one side) and Microsoft/Sun dumbly continuing to ignore Internet in their respective "office" groups.

That's why I suggested using FreeNX and a bootable flash drive, then you DO have all the power on the back-end and aren't limited to a browser on the front end.