new xscreensaver

xscreensaver 4.07 out now...

Tags: , , ,

13 Responses:

  1. ralesk says:

       Barcode!!! :D

  2. dingodonkey says:

    Just curious, does the barcode hack produce an actual, valid barcode, or just random lines that look like barcodes?

    • jtl says:

      It looks like it creates valid UPC E, EAN-8, UPC A and EAN 13 barcodes, but it makes no effort to create actually- or potentially- existant codes (ie, using real manufacturer codes, or doing anything with the pseudo-ISBNs in bookland ('country' 978 EAN 13) barcodes).

      Would a combined webcollage + barcode, that tried to find a book cover or product picture or whatnot, be of interest?

      • dingodonkey says:

        Hmm... I replied, but due to livejournal's pervading awesomeness, I don't see my reply anywhere, so here it goes again:

        I checked the code after posting that comment, and was coming back to retract my question when I saw your reply.

        How exactly would a combined webcollage/barcode work? I just can't picture it.

        • jtl says:

          And I posted another comment which also disappeared -- that one was an out-loud wondering about interest in a PostNet mode for the barcode hack, maybe combined with a MapQuest lookup to show a map of the area/address the barcode goes with.

          My idea for a combined barcode/webcollage goes a little something like this:
          look up the UPC in one of the online UPC databases such as (unfortunately there is no terribly good database available -- the only people with good ones are the grocery chains, and they treat them as trade secrets). After looking up, say, "04968502" and finding out that it's "Coke Diet Cola 12 Oz Single 12 Oz. Can - 6 Pk Can", do a google/altavista/whatever search on those terms, looking for images. Put the found images up on the screen along with the barcode.

          If the barcode is an EAN-13 '978' (bookland) code corresponding to an ISBN, instead of doing a UPC lookup, try to get a cover image from Amazon or

          Of course, doing either of those requires selecting numbers that actually are likely to correspond to something, not simply randomly choosing digits like the current mode.

  3. aaronsw says:

    I was wondering when we'd see a clone of Mac OS X's Flurry.

    • forthdude says:

      Actually, I think this pretty much a direct port of Calum Robinson's OSX 'Flurry' to an xscreensaver hack.

      Of course, I could be wrong...

  4. jfpoole says:

    I tried building xscreensaver with the latest and greatest Apple developer tools (December 2002), and it caused an internal error in GCC. I'm pretty sure this is Apple's problem (since GCC shouldn't crash), but I thought I'd give you a heads up.

  5. baconmonkey says:

    so here's an unfinished thought for ya:

    what about a screen saver that does *something* based upon some manipulation of data polled from the keyboard/mouse when someone was using the machine? obviously for the sake fo security, you'd only store a small chunk of data, and would use the data stream to keem mangling the buffer. i.e. you store 1K worth of input, and rather than overwriting the buffer with new data, you merge it via average, xor, bitwise-or, etc. now what this magic string of data does? no idea off hand. you'd want something more interesting than it being a mere random seed. maybe something like you buffer the mouse movements, and use them as a path to draw something where the keyboard buffer affects the shapes/colors drawn along that path. or even more privacy-centric, you only pay attention to the number of keystrokes that occured every 20 seconds, and ignore stretches of inactivity of 40+ seconds. that gives you masically a motion curve of some sort, which maybe controls palette rotation or some sort of sequence of shapes drawn.
    Or what about parsing browser history and mangling data from there to generate soemthing.

    like I said, incomplete idea. essense of idea: make use of some sort of averaging of some date culled from usage to affect the output of the screensaver, without exposing the information that was gathered.

    maybe using weird mined data to affect probablilities of something occuring - you move a point around on the screen, and the probability of it drawing a square is based on the percentage of .mil sites visited, while the probability of it drawing a circle is based on the percentage of .edu sides visited. the color of the shape is related to maybe the lengths of the paths on .com addresses, or maybe from a histogram of how many hits each .com domain got. like maybe there are 20 sites that got one hit, and 10 sites that got 30 hits, and 2 sites that got 100 hits... but you'd have to make it smart enough to scale, and not use arbitrary things like always .com so the squares are based not on .edu specifically, but maybe the 4th most accessed domain or random if there isn't enough usable data.

    I've got some vague idea about using some sort of mined usage data in conjunction with a genetic algorithm, but to do what? and what do you deem a success?

    This is probably one of those things that will either end up almost always looking the same, or be no different than using a random number generator.

    a simle one would be a markov storyteller that builds it's library from cached html files or by loading data from pages listed in the history.

    granted at the DNA, the story would go something like

    "Home | Site Map | News | Help | Logout Calendar Directions Contacting Us
    Flyer Archive Photo Gallery Merchandise"

    ack, I can't stop thinking about this - a screensaver that presents a meta rendition of usage...

    • jwz says:

      During the screensaver exhibit at SFMOMA a couple years ago (they didn't call it that, of course) there was this thing that was the first derivative of a newspaper: they tracked eye motion when people read a paper, then printed a new paper that was just a plot of the eye-trail. It was really neat, you could tell where the ads and headlines were, and sometimes where people were in pictures.