Art Is Not A Media

Don’t re-invent the wheel ....  Something you are taught all your life when you work in software development ..  For tumblripper i wrote my own “update manager” . Big word for a webrequest and a comparison followed by a popup. Recently when downloading a software (terribly sorry, i cannot remember what it was), i saw some files go thru with a strange filename, i went to check it out, and found SQUIRREL . An update manager that kinda works like Microsoft’s Click-Once (which was a good idea, but never really worked that well for me). It’s pretty much the same approach, you don’t need Administrative privileges to install software ( as it goes in your appData directory), and setup is straightforward (actually, it doesn’t ask you anything, you could call it a silent install). I’m pretty happy with it so far, even if latest update (1.5) just brakes Visual Studio integration (and honestly they should have a VS plugin anyways instead of having to manually edit your project files).

When i first wrote Tumblripper, it was for my own usage. At the time i didn't have 4G on my smartphone, and i looked at different tumblr's for photographic inspiration. So i quickly made a small ripper. Couple of months later, I decided to polish it up a bit and release it. I think i just posted it on CNET and one or two other sites. To my surprise it started popping up a lot in google searches, with couple of blog posts about it, and guides on how to use it. If i recall correctly, there has always been a DONATE button on the site (because i already had it made for chrome/firefox extension), but it never got any action. Bug reports and feature requests started rolling it, so i thought i might as well ask people to donate to motivate me on updating this piece of software, so i made an updated version, which could only  be downloaded if you had donated (you'd get a special link). Let's be honest here, the old version worked fine, the new one was maybe 20% faster and had a couple of options more, that's it, let's call it an "Advanced" version. Donations started coming in. Most were at the minimum amount (classy), and some people even managed bypassing the minimum and sending even less (very classy, those had their licence cancelled though). Over the following month's i increased the minimum from 1euro to 2euro to now 5euro. Simply because the PayPal fee's are almost the samen and the software has become much more complex.

Anyways, that's just a brief history, but what i wanted to talk about, is the principle of Donationware. Personnally, i like it . Usually it's for completely free software, and you give a little something if you feel like you should contribute. (I do for a lot of free software i use). In Tumblripper however, i decided to lock part of the software for supporters. I wish i didn't have to, but otherwise i'd get nothing back. Most donations coming in have the minimum amount, but sometimes, sometimes, i get a different amount, and that makes me smile. Let me explain, On my site, to donate, you have to select the amount between 5 and 20€. For every amount in between those values, i have set a little text which basically explains to what you are contributing in my life . What it says when you select a different amount, even just a euro more, it's like sending a hug, saying you care , and it makes me smile for the rest of the day.

Working on the new release as we speak. It’s indexing a massive Tumblr blog, so taking the time to write a little post on the new indexing engine i designed.

Basically, indexing a Tumblr blog takes a while, because you need to crawl through the archive and identify every post.

Here’s how it currently works : 

load /archive page on the tumblr blog

scan tags for posts and add to a list

find the ‘next’ button and the time-stamp it contains

rinse and repeat until nothing is left.

This is basically the equivalent of pulling a thread from a yarn ball. So we can only scan one page at a time. This method has been used in TR since the beginning, and been highly optimised to make it as fast as possible. (Even in V2 having the downloading occur during the scanning, which results in too many I/O errors due to network resource starvation). So for v3, i decided to get back to indexing first, downloading after. But that means waiting sometimes a long time before any downloading gets done ... But .... I had a brilliant idea to speed up the indexing ! Here’s how it now works :

load /archive page on the tumblr blog

scan the calendar widget to determine date of first post

Delta the latest post and earliest post and divide by X

Launch X crawlers at different dates and have the crawler stop when it find a duplicate.

Working until 1 am again.

I need to get this thing out. It seems like every time i open it, it’s almost ready for beta release, and then i fix this and that, and make a better window for the update, and then a cleaner settings window, and ...

But last night’s session was important. I decided to ditch my “update system” (big word for a simple API call and compare the version number), and use Squirrel .   It’s basically something quite similar to Microsoft’s ClickOnce (which in theory was great, but then when you actually use it ... *sigh*). It’s not too bad to set-up, i managed integrating it in my build process so it’s pretty much automated. 

So far it seems pretty cool. I’m just a little disappointed with a couple of points,

Would have been nice to have a web-installer

The loading animation during setup should show always. (It only shows if setup is taking “too long”)

Clearly missing a updater.IsUpdateAvailable() method. I don’t want the software to invisibly update, but i feel like the user HAS to agree to update (preferably with a change log window), like FileZilla does.

On their site, they claim they support “Channels” a-la-chrome (dev/stable/..)  . But this is not the case ! Not out of the box. They probably means that it’s a pattern that is possible with Squirrel. But then they should do a quick guide on how to achieve this.  I would very much like my users to simply be able to select the channel in the “Advanced Settings” tab, and participate in the beta program that way.

They implemented a nice setting that allows you to specify the percentage of rollouts for a Release. But they don’t explain how this is achieved. (Is it a dice roll every time you look for an update ? )

So that’s for the new Update system. It should make updating smoother, and allow for more frequent updates.

I’ve also decided to implement a “Anonymous Usage” box, that will be always ticked for beta testers and free users, and optional for supporters. 

Still working on TR v3.

Things are starting to look good. It’s been so many times i rewrote this thing, but this time i think i took the right road (by basically not trying to cut corners)

I always wanted TumblRipper to be multi-platform and to have a CLI interface. But making multi-platform apps, it’s not that easy. In theory it is, and making some Proof Of Concepts run on multiple OS is trivial. The Mono project is very usable, and gtk# pretty complete too.

But making a GOOD application work in different environments is another question. You need to start making conditions and test for the different platforms, do this, do that.... And because the main user base, is Windows, it’s a shame to not be able to use WPF (which i personally love), and limit to gtk# 

Long story short, i rewrote TRv3 and started by using a correct project structure. Creating different projects for different concerns. Also i wanted a plugin interface, so that made a couple of new projects too.

So now, i have 

Core Project

Plugin Interface ( so that others can build plugins)

Tumblr Plugin (and others to come)

WPF GUI Project

Putting some finishing touches to the first short-film produced with my pal bollzy under the name MicroMetrage ... 

Except we kinda got carried away, and didn't realize how tough the task would be :) Basically we got around 90minutes of video ... so yeah, thats a full film ^^'

But it's important after a hard task to reflect back upon what was just done (yeah, long rendering time, so i'm trying to unwind a bit, blogging ... )

Script ... It's important ... Even though i did my best to plan out before each day of shoot, i realized something .... I couldn't imagine what a "conversation" looks like, especially when you have more than 2 actors talking together .. Basically since there's so much dialogue involved, and usually 3 or more actors talking together , i was a bit lost on how to frame that, so planning those were a pain .. Basically i'd go thru the scene, list the shots i wanted for style, then just improvise on set the conversation shots .... Thats where my second lesson comes in ... Don't over-estimate your productivity ... If you manage to film more than 10minutes of finalized content in a day you're doing it wrong ... And here, i'm almost ashamed to say it, we managed something around 20minutes a "day"  ..... (a day being a very large word, considering the last person arrives on set usually an hour after the said time), they usually take a lunch break and/or cigarette break only once everyone is here, and sometimes they tell you they have to leave early ... third lesson ... Acting is complicated. Make sure you make your actors rehearse the scene before filming, and when they say they are ready, make them do it again and just watch them, make sure their patterns are locked in and they say/do things on the same marks .... LOTS of hair pulling afterwards in editing if the scene evolves while filming ... ( this can be avoided if you can spare a script that will make sure things are perfect). fourth lesson.... Filming with one camera is the old-way and i think the best way ... Sure filming a conversation with two cameras is easier and MUCH faster ... But you WILL run into the dilemma  of camera placement, and compromising your framing to avoid running into the other camera ... Although if like us you are running out of time, and not sure your actors can maintain perfect continuity , then two cams is heaven ... fifth lesson ... Do your dailies .... Sure you just got back home, unloaded the car and filled the elevator with all your hardware on your own and with a broken hand , you just want to crawl in a hot bath with some pinkfloyd and candles (yeah i do that) ... But try to do your dailies asap ..  name your shots, etc... Its a shitty thing to do, but easier if you do them one by one rather than all of them once the filming is over :) Also it can help identify problems before it's too late .

sixth lesson, give a camera to someone and have them make a behind the scenes, because its fun, and you'll regret not having one afterwards. (try to get someone who has their own cam, and knows how to do some basic editing, so you don't have to worry about that ^^)

So, 

as you've all heard, the Playstation Network was recently hacked, leading to thousands of personal information including Credit Card numbers to fall into wrong hands.

Now we all like that usefull feature to just be able to buy a 5€ game without opening the wallet and having to enter our CC info for an Nth time .. 

But on the other hand, it does mean that Sony (and now some hacker groups) has all the data needed to bill us....

I've been thinking a bit about this issue recently, and in more general terms about database encryption.

Would it be crazy to implement a per-row encryption based on a user key ?

Let me explain .. You'd enter your personal information on a secured website. Along your info is an encryption field, where you enter the password you want.

All your personal information will then get crypted using that password, and that password will NOT be stored in the database. If you lose it, you just reset your personal information, and start again.

For this to work, you would need two things :  * Symmetrical Encryption * No traces of the user-key or a checksum allowing to test if the data has been correctly decrypted. So basically, we'd push the personal data as an auto-fill on a "BUY NOW" field. You'd check if the fields are all okay, if not you retry your password, or start blank.

The system is off-course hackable (what isn't), but would reduce the number of victims significantly, since the cracking has to be done NUMROWS time, and cracking a row would need to use a dictionnary to try to guess if the data has been well decrypted.