Lightning Ray

I made an account here about a year ago, and honestly, the reason I made an account is not because I wanted to post in the forums or such, but to take advantage of "My Follow" feature. This really helps me in keeping checks on the latest episodes of my followed comics, because as a member of the workforce, I rarely have the time to scroll down, look carefully for each pages on what's new and etcetera. In fact, before I started using this feature, I often forgot which comics I follow, so when I remembered at times and tried to check the updates, I don't know where to continue.

I think we could use features like this to "bait" users to sign as members. Think of it like the relation between g.e-hentai.org and exhentai.org (I'm really sorry for using porn sites as reference, but I can't think of any better example). Normal users, or guest users can browse g.e-hentai gallery freely without restriction, while in other hands exhentai gallery has some contents that are not uploaded in g.e-hentai and you have to be a member if you want to see this restricted content.

While it might not be feasible to do content restriction in all-ages comics, unlike 18+ ones (cause I know those "secret" contents are actually somewhat disturbing porn), we could replace "content restriction" with "members-only special features". People basically hated being ordered, and to prevent that, we need to go around and strike from the other side, which in this case we don't actually restrict the content, but we give them lots of beneficial features if they become a member. Or at least that's my opinion, according to my own experience with this site.

Software will not do the trick here.

Basically, to shield a site from DDOS, you would need computing power that can keep up with the incoming information.

And since you are fighting against a bot-net, most of the time, you are talking google-datacenter size here.

Software might help up to a certain point, but in the end it's a resource management problem. You're having -let's say- some 200.000 computers, starting up -again, let's say- 5 connections each, which will be kept alive for about five seconds. so, in those five seconds you have a million threads (1.000.000)  running on your server, and each of these threads has to be checked.

Now in programming terms, that's not TOO bad, but this is where it gets tricky:

While the server is trying to handle the 1.000.000 requests, the botnet will just send NEW requests, adding to the pile, and stuffing the server's memory with unneeded threads.

If you are running a Linux server you might make a save, by using the SWAP space, but that's slow, making the process of eliminating requests slower too.

Remember, the server has to check everything, while the botnet does not. it just blindly keeps opening new connections.

In algorythmic terms, I suppose you could say that the server tries to deal with the data in real time at a speed of O(n) which is the same as the mathematical "y = x"

(in case of multiple servers working it might be divided, 100 servers make O(0.01n) or "0.01x = y")

While the botnet can keep opening connections in realtime with just one instruction O(1) which is the same as "y = 1"

That way you can see it's always an unfair battle, because for every step the botnet does (1), the servers have to work through a list of connections (n/(amount of servers)). They will always stop at some point.

That's a lot of madness to deal with.

The only servers heavy enough to deal with that, are the ones dealing with such traffic everyday. These are server parks capable of running as top-level DNS server. A.K.A: Google, NSA, IBM, and all those other companies I would love to have a look inside XD

I suppose cloudflare has some of those server-parks as well, but seemingly that was not to grumpy's wishes. (which might mean that they do not actually have the right equipment)