tain.totalcodex.net

vinylrake wrote:How about setting up a single Myth server signon site where one could register for a new account that would work on either server? The single signon account registration site could incorporate both server data requirements (e.g. no free email accounts, etc) so the requests that made it past the data validation screen would be 'clean' and ready to add to the individual server user-account table(s).

New account registration requests could be distributed to the game servers on a regular daily/hourly/whatever schedule pretty easily, and adding the requests to the account db could be automated fairly easily (I imagine) if the server admins were open to the idea. (e.g. the new accounts could be sent as formatted SQL insert statements formatted to match each server's account file structure).

Why create unwieldy and difficult solutions to a problem that is very easy to solve?

Death's Avatar wrote:

vinylrake wrote:How about setting up a single Myth server signon site where one could register for a new account that would work on either server? The single signon account registration site could incorporate both server data requirements (e.g. no free email accounts, etc) so the requests that made it past the data validation screen would be 'clean' and ready to add to the individual server user-account table(s).

New account registration requests could be distributed to the game servers on a regular daily/hourly/whatever schedule pretty easily, and adding the requests to the account db could be automated fairly easily (I imagine) if the server admins were open to the idea. (e.g. the new accounts could be sent as formatted SQL insert statements formatted to match each server's account file structure).

Why create unwieldy and difficult solutions to a problem that is very easy to solve?

If the solution is to kill one of the two privately run servers that's not a good solution in my opinion.

What Myth really needs is Jabber-style decentralized identification. As it stands, either you have multiple independent servers (requiring people to maintain separate logins for each) or you have a single server (rendering it totally nonredundant—even as it stands the Myth metaserver community hangs by a thread). Actually, the real solution is to do a giant overhaul of the entire metaserver concept, but, well… I think the obstacles there are obvious.

~J

Well, I don't necessarily find the idea of sharing accounts from a Central Authority (CA) completely repulsive, something like that should work quite easily ...but formatted SQL (IMO) is a horrible idea for a bunch of reasons, hehe.... given the nature of an account record, a simple delimited line in a text file is all that is needed and let the processing of it happen in whichever way each server chooses to do so. Updates don't even need to happen hourly or anything like that, only when there are new accounts. I also would not have the servers retrieve the info unless the location they retrieve from is closely guarded and firewalled to non-metaserver IP's. I would go the other route of having the CA upload (perhaps sftp) to designated areas on each server, then require each server to remove that file to another spot on it's system for processing. This will minimize the window in which that data could be intercepted (I'm assuming it will be passing, names, email addresses, usernames and whatever form of registration code is decided upon). You also have to bear in mind that whatever would be chosen, it must be constructed in a manner that it will work with both servers' existing means of account creation and you will have to recreate the security enjoyed currently in that all accounts are local, for both metaserver and web, to the respective metaservers. I sure won't be accepting data from any place I don't trust 1001% and then some.

As for the metaserver thing hanging precariously... all I'll say is that it is only as precarious as you want it to be. The 'Spaghetti Code' is out there, the Mariusnet metaserver is freely available (a pretty well documented and already compiled, ready to run out-of-the-box metaserver for the Win32 environment), Blades released his code I believe and then there is the updated source that someone on here released (sorry, forgetting names on who did it ).
If you think another server is needed, then pick one and fire it up.

I can definitely understand the desire for a CA however.

Considering PMnet can't even get normal user account registrations going right now, nor guest access, I don't see them supporting CA either... They need to get their stuff together, or shut down so people can play on Mnet with new players.

PrplPplEater wrote:If you think another server is needed, then pick one and fire it up.

Two words: hosting cost. I could put a metaserver up… well, not tomorrow (I could if I tried hard enough), but the next time I'm home.

I could not, however, guarantee even one nine of uptime. Or performance—it's on a residential-grade internet connection (fortunately, with an ISP that leases fixed IP addresses and permits servers in their TOS). I can't justify the money for colocation, and application hosts aren't cheap the way webhosts are these days.

~J

For 7 years MythForums.com, and for the last 5 years Mariusnet... were hosted out of my house on a DSL line with only 768k upload. It is only recently that I moved everything to a true hosted dedicated server. Up to this point, it was on home built machines, on battery backups bought out of pocket and all of it run on my DSL line.

That machine was running MF.com, Mnet.com, the metaserver, a couple more websites, a Halo Dedicated server and a Neverwinter Nights server with a custom module with over 300 areas. Toward the end of last year, it also hosted what turned into a high traffic website (20k-30k unique page views per day) and it was only then that I started to run out of bandwidth.

Yeah, my reliability wasn't 99.9%, but it wasn't horrid either except during the times when a drive died or power supply blew... or someone would be downloading crap from MF.com making me lag like hell when I played Medal of Honor.

The bandwidth required of a metaserver can realistically be run on a 128k upload (actually, it can be run on a dial-up connection, but we won't go there). If you really wanted to offer a good environment though, you can do it on the cheap by getting a 2nd DSL line or just renting a VPS for like $50/month from someplace like http://www.asmallorange.com .

If you want it, the cost won't be an issue

and what kind of computer is needed to host the metaserver ?

I ran all of the above on a dual processor Athlon MP2200+ system with 1GB ram.... but Anything running WindowsXP will run the Mariusnet app just fine... any comparable machine runnning any flavor of Linux will run the others.

Technically speaking however, an old ass 486 would be sufficient (and I'm pretty certain that is what Bungie.net was run on). When you get right down to it, the metaserver really is nothing but a chat server that does authentication and then lets other people find games... it doesn't actually host any games at all. It's the lowest of overheads.

PrplPplEater wrote:...but formatted SQL (IMO) is a horrible idea for a bunch of reasons, hehe.... given the nature of an account record, a simple delimited line in a text file is all that is needed and let the processing of it happen in whichever way each server chooses to do so. Updates don't even need to happen hourly or anything like that, only when there are new accounts. I also would not have the servers retrieve the info unless the location they retrieve from is closely guarded and firewalled to non-metaserver IP's. I would go the other route of having the CA upload (perhaps sftp) to designated areas on each server, then require each server to remove that file to another spot on it's system for processing. This will minimize the window in which that data could be intercepted (I'm assuming it will be passing, names, email addresses, usernames and whatever form of registration code is decided upon). You also have to bear in mind that whatever would be chosen, it must be constructed in a manner that it will work with both servers' existing means of account creation and you will have to recreate the security enjoyed currently in that all accounts are local, for both metaserver and web, to the respective metaservers. I sure won't be accepting data from any place I don't trust 1001% and then some.

For security and good process/data flow/design I definitely agree, I was positing a lowest common denominator type solution demonstrating how simple the process *could* be. The problem of course with a single Central Authority for new user registrations is that it once again creates a single point of failure for the whole process - in this case for BOTH servers. Ideally I think the best solution would be to have the "New User Registration" signup process be decentralized. Possibly by creating a standard 'canned' webpage with builtin hashing (or PGP) that would encrypt the user registration data and send it off to the game servers' "in boxes". That way anyone that wanted to could add the "Myth Game Server - New User Registration" page to their site. Worst case scenario would be you would get banned people adding the page to their site to create bogus accounts, but I would imagine the standard IP/site blocking that is (I assume) in place for the current new user registration process would handle this.

OK, I take it back - the BEST solution would be to have an application that listed games and players currently online on all Myth game servers so a player could drop back to the starting screen and switch servers.

Actually the absolutely best solution would be for Myth to handle multiple servers as if they were seperate rooms.

a standard 'server finder' like you see in *EVERY* game out there nowadays would definitely be ideal.

The flip side to that is that there are barely enough people to warrant more than one server and we're discussing allowing for the addition of as many as the proposed architecture can handle. It's a serious dillution of available players.

I can say for certain that the Mariusnet server is not going anywhere now.
Until I got this new server, I was giving strong consideration to shutting it down because, frankly, I was tired of running things out of my house that prevented me from enjoying my own internet connection. Now that I have something, paid for by that high volume site, that solves all of those problems... I can say for certain that Mariusnet will be around for many, many years to come if for no other reason than the infrastructure it lives in/on isn't going anywhere and it takes no appreciable resources to run it.

As for folks wanting a pretty website... yeah, whatever :p
It's in the plans, I just need to find the time.
I'm kinda leaning toward merging MythForums and Mariusnet's sites into one and killing off MythForums (or leaving it as a sort of mirror).

Myrd wrote:Considering PMnet can't even get normal user account registrations going right now, nor guest access, I don't see them supporting CA either... They need to get their stuff together, or shut down so people can play on Mnet with new players.

Ok lets say theorically the PMnet server is shut down forever.

What happens if Mariusnet suffers a major problem that knocks the server down. Where do people go then?

Same problem happens if we close Mariusnet. As soon as we are left with 1 server and it is taken offline for anywhere up to a week, the Community is effectively dead.

It's not like re-building a site and server system is something that can be done in a week, or even a single month. Especially with RL issues, and the divorce Blades is going through.

And anyways it's not exactly like Myth gets hundreds of new players a week is it.

Valid points... all I can say is that Mariusnet won't be closing. The server it lives on now is paid for by other means on that server, so it's enjoying a permanent free ride. It uses only ~34MB ram and near zero CPU and bandwidth..... it's not going anywhere.

So, at least you can relax a little bit

Horus wrote: It's not like re-building a site and server system is something that can be done in a week.

Honestly, given a week of full time work, I am confident I can set up a new metaserver based on the updated metaserver code, and make a basic site where users can create accounts and reset their passwords & a few more features (rank would take some time to tweak, so that it would match what people expect...).

Myrd wrote:

Horus wrote: It's not like re-building a site and server system is something that can be done in a week.

Honestly, given a week of full time work, I am confident I can set up a new metaserver based on the updated metaserver code, and make a basic site where users can create accounts and reset their passwords & a few more features (rank would take some time to tweak, so that it would match what people expect...).

But Myrd you are a MythProgrammingGod (or at least demi-god?), the average Myth player or script kiddie isn't going to be getting a server and registration site up and running in anywhere near a week.