Safari 4: Caching Images of Every Page You Visit, Where it’s Hard For You To Find Them

Safari 4 beta leaves data, privacy trail in its wake – MAC.BLORGE.

Yesterday I decided to give Safari 4 Beta another try.  It’s supposed to be super fast and all, and I was thinking about those sites like Facebook and which seem to cause a ton of extra CPU usage when I leave them open… thinking maybe a more efficient Javascript engine would make them more pleasant to have around.

I tried it, and it was super fast, but if anything it achieved that speed by causing even worse CPU churn, so after a while oohing and aahing at its speed I toasted it, uninstalled and went back to Firefox 3.

Then I read this article, and the article it links to, and sure enough, in the short part of the evening I was using Safari 4 I had generated 170+ megs of data in a hidden, can’t-reach-without-command-line-fu location on my computer, not even in my personal user directory.  (I don’t think it’s hidden for nefarious reasons; Apple doesn’t roll like that.  I think it’s hidden because this cache is what the Apple engineers needed to achieve the effect they wanted, and they didn’t think you should have to worry your pretty little head about how they did it or how much of your disk space they used to do it.)

And darned if that hidden cache directory didn’t include full size image files of every site I’d visited during that time.  Which stayed behind after I uninstalled the beta.  How much data would there have been after a week of usage?  A month?

Lame, Apple.  Lame.

Clipperz review

A while back I happened across this article on — “Moving the ‘C’ in ‘MVC'” — about web applications whose logic is implemented primarily on the client-side, in javascript, as opposed to the server-side. Interesting stuff, the sort of thing that’d bubbled around my head at one time or another.

That article mentioned and their concept of a “zero knowledge web app.” The idea here is to implement client-side cryptography in javascript and store nothing on the serverside except wads of encrypted data. Data that is encrypted on the client, that is. So it never passes over the wire in the clear, and it’s never decrypted on the other side of the wire.

That’s pretty cool. In fact, hosts their own best shot at a zero-knowledge web app, a password manager. A web-based password manager seems insanely insecure, but not so if you implement it as a zero-knowledge web app. It’s basically like keeping a local HTML with all your passwords and javascript which allows you to click on those passwords and have it launch the appropriate page and log in to the service in question (Google, Facebook, whatever) — except that instead of keeping it locally you’re keeping an encrypted version of it on’s servers.

As far as usage goes, it’s pretty simple, but unusual. In a normal web app, you log in once, after which you are recognized by cookies until some timeout period. That’s because you’re logging in to an app running on the server, and cookies help maintain the illusion of a persistent connection. In clipperz, the app is running in your browser. So if you close the page and come back to it, you have to log in again. But you can leave that page open forever and it never times out, because there’s a real connection between you and the app (an open browser window) so there’s no need for the whole cookie deal.

Adding logins to the app is an unusual process but easy once you get used to it. When you’re on the login page to the service in question (myspace, yahoo, whatever) with your username and password typed in the blanks, you can click on a bookmarklet you’ve previously saved from clipperz, and it will extract a chunk of JSON data from the page you’re on, representing the login form, your username, and password. You cut and paste that into your running clipperz session in another tab or window, and it takes that JSON chunk and adds that to its data store, and now you have a clickable link on your clipperz page which will log you in to that service.

Knowing I’ve got an instant one-click login to a service makes it easier for me to make a habit of logging out of a service when I’m not using it.

It’s also easy to maintain more than one login to the same service using clipperz. Each is just one click away.

Clipperz has a sometimes-friendly, sometimes not-so-friendly competitor named “PassPack,” whose authors consider’s zero-knowledge thing “fallacious” in some way, though reading through it I’m still not sure exactly what they think is wrong with Clipperz’s way of doing things. I guess the idea is that any privacy and security that Clipperz provides for you that PassPack doesn’t is just silly and you shouldn’t worry your pretty little head about it, for such worrying is a “fallacy.” I don’t know. It sounds like PassPack might be easier to use, and harder to understand what they’re doing. With clipperz you just have a username and passphrase; with PassPack you have a username and password and an additional crypto key. (UPDATE: see comments below, from Tara from PassPack for clarification of what they were getting at.)

In any case, PassPack and Clipperz have each posted a list of the pros and cons with their respective services. Overall I have to say that having actually used Clipperz for a while I don’t see anything about PassPack that encourages me to explore it as an alternative.

Anyway, my review of Clipperz after a couple weeks of use — thumbs up! It’s unusual but worth getting to know.

A Tasty Video Recipe

I like to download videos from youtube using the magically delicious DownloadHelper. Especially music videos. The only sad thing is that the video quality tends to be poor, and it can only get worse if you convert it to a less crappy format than the standard flv — such as mp4 for example.

But Youtube is being kind to us. Many, many videos on youtube, if not all, are available in a higher quality format — in fact, mp4, the format that is standard for things like iPods. The magic you do to get the higher quality is take the youtube url and add “&fmt=18” to it. That’s all. Just add that to the address in the address bar and reload and you will get a higher quality video, which you can then download with downloadhelper. You might need to change the .flv extension to .mp4 to play it back with Quicktime.

There’s also a thing in Youtube’s account preferences you can set to get higher quality stuff by default, but I don’t think it’s the same — I think (not sure) it gives you the fmt=6 version. (there are at least two different higher quality settings, &fmt=6 and &fmt=18 — I think the &fmt=6 is better than the default but not mp4-encoded like &fmt=18.)

There’s a greasemonkey userscript to do the fmt=18 for you.

UPDATE: if you see something in the menu called Video.mp4 that’s guaranteed to be the good stuff. I think it also shows up under other names sometimes too — downloadhelper sometimes offers several links to the same video so it’s confusing.

UPDATE: it ain’t just video. The audio is *noticeably* better in the higher quality downloads. Much crisper, makes the old kind sound muddy in comparision.