CWE-ou & Minkata Road Map Planning

[JWPlatt]

FYI - Cyan's iPad Riven project seems to be nearing release, and Mark would prefer to do this update before the concurrent shutdown of MQO on December 31, 2012, and implementing the proposed annual AWS contract for MOULa on a reserved instance. We should be prepared for an imminent update, so your timing is good. Keep in mind that Mark has some library work to do with this update as well, per your instructions. I have passed them on to Mark, and I think you did as well. He'll probably need a reminder about what he's getting into with this one.

I'm really looking forward to all those changes. It's a nice list.

That said, all looks good with the exception of chat scrolling and login credentials. I have not seen chat scrolling tested. If you are validating it, and it's small (a few lines of code in one or two files), that seems okay for inclusion. It would be better if it got some testing by several people on Minkata. I am aware that Mac_Fife has modeled the credential fix, but not in c++ code for the client, so I doubt that will be ready. I was not at the AGM, so maybe there's more up to date info? But my guess is the update after this one for that.

[rarified]

I had mentioned the login credentials, partly because I was bitten by it trying to get into AGM

The base C++ code is straightforward, but at the end of our last email discussion we were trying to decide how to avoid invalidating the entire MOULa user base with the "fix". I had suggested implementing a heuristic to try to be backwards compatible, perhaps by trying the "new" seeding first, and if that failed, try the "original" seed for creds. If the second method worked, convert the local copy of the login creds to the new seed and rewrite the authentication file.

But AFAIK we didn't come to a consensus about that. Perhaps a working prototype of the compatibility code is in order?

_R

[Christian Walther]

Yes, of course things need to be tested before going to Cyan. I will not merge things to CWE-ou that haven’t been on Minkata for at least a few days. (We did that once with the Bahro stone sharing stuff, and came to the conclusion that we were not going to do it again.)

If the MOULa update is imminent, then I think we should merge the current Minkata state to CWE-ou right now, not including any of the untested changes, to make sure that at least that makes it into the update. Unless Skoader has any physics updates that he wants to get in before it goes to Cyan. Then open a new testing cycle ASAP with the security fix, cursor update, and possibly chat scrolling, and hope that we’ll be able to complete it before the update too. Anything further, particularly stuff that hasn’t even been coded yet, gets pushed back to the cycle after that.

Here’s a couple of scenarios that I can think of:

Code: Select all

| = cycle boundaries
^ = MOULa update
(last) = summaric listing of the changes of the last cycle that are not on MOULa
yet, currently listed on
http://wiki.openuru.org/index.php?title=Minkata_Shard:_Contributions_under_Test

| (last) | phys steal exec curs scroll login |      - unrealistic if the MOULa
^                                            ^        update should rather come
                                                      sooner than later

| (last) | phys steal exec curs | scroll login |    - good, but still risky - 
^                               ^                     risk is falling back to…

| (last) | phys steal exec curs | scroll login |    - … which is probably not
^        ^                                            what we want

| (last) | phys steal | exec curs scroll login |    - unfortunate because we
^                     ^                               would miss exec and curs
                                                      in the MOULa update

| (last) | phys steal | exec curs | scroll login |  - good
^                                 ^

| (last) | phys steal | exec curs scroll | login |  - good
^                                        ^

[Skoader]

I'm happy for the physics changes to go to Cyan as they stand. While there is still work to do, what we have now is already a significant improvement.

I was going to take a deeper look at the chat scrolling fix this weekend (at a cursory glance I had some concerns about redundant scrolling), but I've run short on time.

[Mac_Fife]

I had mentioned the login credentials, partly because I was bitten by it trying to get into AGM

The fix for the cached login credentials is easy enough (I have that) but there are a couple of snags I'm having: I think there's a difference in MSVC includes between 2003 and 2010 versions (that I haven't been able to confirm yet, since I was having trouble trying to get the 2010 version installed - [don't ask, long story]) - That means there's probably a better (more correct and portable) way of doing things than I've done. Secondly, I was trying to build a checker/patcher for the login.dat file and I'm clearly missing something important since I'm not getting the cryptKey to match for the old version .