Page banners and other things of interest.
POSTED BY: HAKEN
UPDATED: Thursday, October 28, 2010 15:36
VIEWED: 10507
PAGE 3 of 4
Kwicko wrote:
Quote:
o, you don't. The avatar, I mean. Unless your avatar is supposed to be a small blue box with a question mark in it, in which case you have a shiny new avatar!
You have to look at it on the beta site. We're working through some bugs with it.
http://beta.fireflyfans.net/mthread.aspx?bid=27&tid=45804
Haken
NOTIFY: Y | REPLY | REPLY WITH QUOTE | PERMALINK | TOP | HOME
Sweet.
By the way, did you just tweak something in response to my remarks about Safari and notifications, because it just started working again.
I'm running the latest Snow Leopard and Safari updates, and the notifications and links HAVE worked recently, but just quit the last four or five days or so, but just started working again.
The modern definition of "socialist" is anyone who's winning an argument against a tea-bagger.
AURaptor's Greatest Hits:
Friday, September 24, 2010
I hate Obama's America. You're damn right about that.
Friday, May 28, 2010 - 18:26 To President Obama:
Mr. President, you're a god damn, mother fucking liar.
Fuck you, you cock sucking community activist piece of shit.
... go fuck yourself, Mr. President.
NOTIFY: Y | REPLY | REPLY WITH QUOTE | PERMALINK | TOP | HOME
Nope, not working again.
Quote:
ADODB.Recordset error '800a0cc1'
Item cannot be found in the collection corresponding to the requested name or ordinal.
/mthread.asp, line 118
The modern definition of "socialist" is anyone who's winning an argument against a tea-bagger.
AURaptor's Greatest Hits:
Friday, September 24, 2010
I hate Obama's America. You're damn right about that.
Friday, May 28, 2010 - 18:26 To President Obama:
Mr. President, you're a god damn, mother fucking liar.
Fuck you, you cock sucking community activist piece of shit.
... go fuck yourself, Mr. President.
NOTIFY: Y | REPLY | REPLY WITH QUOTE | PERMALINK | TOP | HOME
Kwicko wrote:
Quote:
By the way, did you just tweak something in response to my remarks about Safari and notifications, because it just started working again.
If you're referring to the beta side, then yeah. I've been doing daily builds to fix things.
NOTIFY: Y | REPLY | REPLY WITH QUOTE | PERMALINK | TOP | HOME
[quote=Haken] That's the normal distribution and the more frequent appearing ones are probably the mean. There's really no way to get around that except for the site to keep track of the ones you've seen and not repeat any until all the banners have been viewed. It's not that difficult to implement, but seems like a waste of resources since the index value will have to be stored as a cookie on the browser end or as a value in the database on the server end.
I have a couple of other things I can try. One is time based where I take the second that the page loads and MOD it with the number of banners, but I'm sure over time, we'll end up with the same normal distribution problem. Just can't really get away from it when we go the random route.
Pretend this is Chinese: << Oh Dear Buddha! >> I wish I had never suggested “random”. Serving banner is definitely a case where NOT to use Mersenne twister pseudorandom numbers. I should have said the simplest thing imaginable: “Serve the banners in order: 1, 2, 3, . . .” If there's only one person clicking very fast “open link in new tab” on fff.net 2.0, there will be no repeats until all the different banners have been served up. And that's good enough for me.
NOTIFY: Y | REPLY | REPLY WITH QUOTE | PERMALINK | TOP | HOME
Buddha is wise for a reason. He sees all things in the universe as a whole. Your order is not the same order as that of the rest of the universe. Your 1 is not my 1. My 2 is not your 2. We do not live alone in the universe.
In order for the site to serve up the banner sequentially, it'll have to store that counter somewhere. And that counter applies not just to you, but for everyone that is currently on the site. You may be at 1, but another who hit the site before you could be at 5, and another could be 26. Depending on the number of visitors concurrently on at the same time, managing and incrementing these counters "might" use up a considerable amount of resources.
Your idea sounds good if it was just for a single person, might even work well for a lot more, but FFF.NET, as small as it is, still gets a couple of thousand visitors a day who view thousands of pages. Just something to keep in mind when you think about this stuff.
Maybe if I had the memory and server resources, it might be possible. Meanwhile, I still think the random route might be the way to go. At least with that, in a given span of time, a site visitors will eventually see all the banners.
NOTIFY: Y | REPLY | REPLY WITH QUOTE | PERMALINK | TOP | HOME
I did not mean many individual counters, each exclusively assigned to a different visitor. That's too many counters to keep track of. I meant one counter serves all visitors. Like monotheism has but one omnipotent God, fireflyfans.net could have one Buddha-counter controlling which banner is served next. Any time that anybody is served a page, Buddha-counter is incremented and the corresponding banner is sent. You only need enough memory for that one counter.
Millions of people could be served by that one counter.
NOTIFY: Y | REPLY | REPLY WITH QUOTE | PERMALINK | TOP | HOME
Hmmmm...okay, think about this for a bit. What increments this counter?
NOTIFY: Y | REPLY | REPLY WITH QUOTE | PERMALINK | TOP | HOME
Quote:I get to be the Fool because I don't know what I'm writing about, but why should I let that stop me?
Haken: What increments this counter?
Increment the counter whenever creating the source of http://beta.fireflyfans.net/mthread.aspx?bid=12&tid=3456 or whatever other page has a banner. (There is only one counter for all pages and all users.) Is the counter greater than the number of banners? If yes, set the counter back to 1.
There is a line in the source, where ## is both the number from the counter and the number of the banner:
I just ran the stupidest possible test by repeated hitting the refresh button for page http://beta.fireflyfans.net/about.aspx I only had to hit refresh twice before the random number generator gave me the same banner again, out of 42 possibilities. Second test: I refresh three times before the random number generator repeats a banner. Third test: I refresh five times before a banner repeats. I don't know that the dumb counter I'm proposing is an improvement. I have suspicions.
For people who just stumbled on this thread -- there is always what was being done last week: the banner changes about every minute, taking better than a half hour to cycle through all 42 banners.
NOTIFY: Y | REPLY | REPLY WITH QUOTE | PERMALINK | TOP | HOME
This is actually a really good statistical problem. We have a couple of options here, all of which will end up with a natural distribution of some kind in terms of which banner is viewed.
Here are the options:
A. Random banners for random viewers
A random banner is displayed for a random user who views the page.
B. Random banner for sequential viewer.
A random banner is displayed for a user who views every page.
C. Sequential banners for random viewers.
Every page is assigned a banner sequentially. Not all users will want to view every page.
D. Sequential banner for sequential viewer.
Every page is assigned a banner sequentially, viewers will view every page.
E. Sequential banner assigned to viewer.
Banners are displayed sequentially for the viewer regardless of page.
I think that's it. It was already decided that while E is probably the best option, but it is also the most resource intensive. Option D would be great, but by human nature alone, that's not going to happen since not everyone will view every page.
That leaves us options A, B, and C. Option C, the natural distribution will go to the users. Option B, natural distribution goes to the banner. And Option A, which is what we're doing now, has natural distribution go to both banner and viewer.
So, statistically, I think option A is best because we have overlapping natural distributions and over time, everyone will eventually see all banners. Of course, the same could be said about B and C, but I think option A is probably the quickest because we get a better distribution.
If there is anything we can tweak here, it's probably the randomness. That fancy algorithm I'm using isn't perfectly random. I've been looking at http://www.random.org and may possibly use that to get my numbers, but that seems almost an overkill for what we're attempting to accomplish.
So, I dunno, maybe just watch the banners for a while and see? If overtime you don't see all the banners, then we'll just make changes at that point.
Regardless, this problem was fun to think about.
NOTIFY: Y | REPLY | REPLY WITH QUOTE | PERMALINK | TOP | HOME