To sum up all the changes in a sentence: AjaxLife is now hosted on Amazon AWS.

In more detail:

AjaxLife’s texture cache is in an S3 bucket
This ensures that it won’t get lost, and also that it works comparatively well.

AjaxLife’s static files are in an S3 bucket
Hopefully more reliable than my hosting. Definitely much faster than my hosting.

AjaxLife itself is run on Amazon’s Elastic Compute Cloud
This means that AjaxLife has a bigger server to run on. It also means that, once I’ve figured out the critical number, I can set up the loading system to bring up another server if needed. Another advantage is that AjaxLife should finally self-correct failures – should the AjaxLife server fail, the server it’s running on will be terminated and a new one launched within ten minutes. The servers are configured to load AjaxLife as they boot up.

ajaxlife.net is now on the same server as katharineberry.co.uk
Saves me £35/$70 per month. The new system actually costs more, but is also much more powerful. More on that later.

As such, logging onto AjaxLife should now follow a sequence something like this:

1. Navigate to ajaxlife.net in your favourite web browser.
2. You will may receive a message telling you AjaxLife’s down, in which case it should be automatically restored in a few minutes. You will otherwise be redirected to a seemingly random IP
3. Log on and use as usual

BIG IMPORTANT NOTE THING: The observant among you may notice that AjaxLife no longer uses HTTPS. Don’t worry – your details are still safe. Since HTTPS is a significant contributor to lag between the client and AjaxLife server, not to mention the impracticality of getting SSL certificates for whatever odd hostname the application ends up on (ec2-67-202-36-23.compute-1.amazonaws.com, anyone?), it was dropped. Instead, AjaxLife uses 1024-bit RSA encryption to ensure that your password cannot be read in transit, combined with a challenge/response to ensure it is not susceptible to replay attacks. Although anything past login is not encrypted, this is not actually a loss – they were never passed to SL encrypted anyway. That said, if it bothers you, try https://old.ajaxlife.net/client/ – although the certificate won’t be valid, and it’ll usually be down.

Other changes made with little or no relation to AWS:

• AjaxLife can now download textures with transparency correctly
• AjaxLife will pull profile images directly out of SL’s search, for speed reasons. This results in a tiny watermark in the bottom-right corner, however.
• The minimap seems to work again
• Updated to libsl 0.3.2

I hope you enjoy the theoretically more reliably service. However, there is another note to add: These improvements cost me money directly proportional to your usage. If you find you use the service frequently, please donate using the button (that will be) in the sidebar. Thanks!

Having poked around with it, I have noticed the following – based on the incomplete version that I found on their site:

• It looks good
• Classifieds are integrated neatly into the system
• A good chunk of the data is hosted on Amazon S3 – profiles, events, places, top traffic sites, etc.
• The TG and MG are in separate S3 buckets
• It appears to use new secondlife:// features – specifically, secondlife://app/event/*/about, secondlife://app/classified/*, secondlife://app/agent/*/about. Visiting these with the RC client open made it crash. That said, clicking most buttons made it crash, so that’s not saying much.
• Teen Grid and Main Grid results are not separated. I assume that this will change, given the separation of their data.
• It searches wiki.secondlife.com.
• A good deal of the data isn’t on S3 yet, so many links just tell you the key couldn’t be found.
• Main Grid profiles default to visible, but Teen Grid profiles default to hidden.
• Parcels have a list of searchable objects on them along with their prices, arranged in alphabetical order. Their parcel image is also visible.
• Boolean operators such as “NOT” work as expected.
• The search does not apply to the owner of something – e.g. a search for “Alex NOT Harbinger” still returns things owned by Alex Harbinger.
• You can now grab limited resolution textures off the web at http://secondlife.com/app/image/texture-key/2 – this is used by the search to provide images.
• The popular places lists, which appear in search (but probably shouldn’t), show sandboxes that are hidden from popular places in the current system.
• Objects that aren’t for sale tend to appear in the object listings. Oops?
• Avatar names tend to be just “(waiting)”, although the profiles they link to are accurate. I have yet to find any marked as “(hippos)”, however.

Also, lots of people seem to have expensive objects called “Object” for sale. It’d be nice if those were filtered out. The information seems to be fairly up-to-date, but is definitely slower at updating than the in-world stuff.

Of course, this is presumably supposed to be unreleased, so it could all change at whatever time. Still fun to look through though. And it actually looks like it’s a significant improvement over what we have now, woo.

Just need some way of separating out the MG and TG results. I suspect that LL will achieve this by locking searches to one of the two buckets depending on which grid you use.