User talk:Serge Yseron/Archive10

Because i like to remember who was there, and you ?

Serge Yseron 18:17, 6 August 2011 (UTC)

We love you too

Sometimes, not always, but sometimes, we dont have 2 leechers and 2 luxons beaters playing as kurzicks in the same battle. It's a good thing that we can play on an even ground, dont you think ?

Serge Yseron 23:54, 6 August 2011 (UTC)

Welcome to the b...

Serge Yseron 19:34, 8 August 2011 (UTC)

Braziiiil, la laa la laa la la laa laaaa...

Serge Yseron 22:40, 8 August 2011 (UTC)

$*<#@ - $*<#@ - $*<#@ - $*<#@ - $*<#@

Best rupter in the world, here. Yseron - Serge Yseron 18:05, 11 August 2011 (UTC)

Visual effects

Aux hommes de bonne volonté... Yseron - 90.15.177.127 18:15, 13 August 2011 (UTC)

Unexpected cooperation

Sometimes, when the matchmaking is kind enough to throw ordinary players at us (and god know the matchmaker hates me), we get the surprise of having the opposite faction kindly giving us the victory...
And yes...our dervish had a banana scythe...

Serge Yseron 21:21, 13 August 2011 (UTC)

World's best definition of bravery

not suitable for children below something around 12. Yseron - 109.213.25.111 12:45, 14 August 2011 (UTC)

Feathers price check

for Chicken Super Armor(TM).Yseron - 109.213.25.111 13:32, 14 August 2011 (UTC)

Logic

Wondering why people seemed to hate dervishes that much in pve, i tested a balanced grp involving a dervish. The thinking that went into it was the folowing:

Primary hint:
- Designers cant make elite areas for a given campaign that would exclude professions tied to this specific campaign. Hence my conviction: there must be a way for a dervish to play there without any drama.

Secondary hints:
- In a video and in some dev posts we can read that devs like dervishes, wich should lead to think that, even if they hadnt made the campaign that gave birth to dervishes to be dervish-friendly, they gave the dervish some advantages of some sort, probably overpowered.
- They decided to increase dervishes armor through mysticism in pve. So it had to have something to do with enchantments that cant be striped. The dervish elite came right after this thought. The result is magnificient as far as the foundery is concerned.

This first try with a melee (i usually never play with one) gave me 2 titan gems and some geenish butter knifes of nihil in the first 15 minutes.

Serge Yseron 13:32, 15 August 2011 (UTC)

there's two problems with dervishes, one is that the hero melee ai sucks, the other is that you could be bringing a mesmer instead, also wtf is a geenish butter knife of nihil 24.130.140.36 19:04, 15 August 2011 (UTC)

To show a screenshot with a dervish in it i cant replace the dervish with a mesmer. A "geenish butter knife of nihil" is "Greenish butter knives of Nihil" spelled Yseron style. Yseron - 90.9.120.133 19:16, 15 August 2011 (UTC)

Dummies revenge

Reverse pure strike. Yseron - 90.9.120.133 19:34, 15 August 2011 (UTC)

Mole X best profile

By the rivers of Elonnaaaa....

Serge Yseron 17:09, 21 August 2011 (UTC)

I am aware that people continue to consider that necros are useless in doa, and just tolerated when playing glaive rit. To read the skills descriptions is the best compliment you can do to your brain.

Serge Yseron 17:58, 21 August 2011 (UTC)

Time has nothing to do with it

It's time for some bad dancing and bad singing again.

Warmup : 1, 2, 3, go !

Time has nothing to do with it - Georges Brassens:

When they are brand-new
just out of the shell
of the cocoon
all young greenhorns
take old geezers
for as many jerks

When they are here
the hoary heads
the grey noggins
all the old fogeys
take young lads
for as many jerks

As I’m caught between two ages
I have for them all a message

[Chorus]
Time has nothing to do with it
When you’re a jerk you’re a jerk
Be you twenty or be you a grandpa
when you’re a jerk you’re a jerk

No more controversy between you
Obsolete jerks or novice jerks
Little jerks of the latest shower
Old jerks of snowfalls of yore
Little jerks of the latest shower
Old jerks of snowfalls of yore

You budding jerks
innocent jerks
young jerks
who, don’t deny it, take all dads for stupid jerks

You aged jerks
worn jerks
old jerks
who, admit it, take all rookies for bloody jerks

Meditate the impartial message
of one who’s caught between two ages

Yseron - 90.15.58.138 22:59, 21 August 2011 (UTC)

Busy minds makes happy lifes

When typing something related to other things i do to entertain myself i suddenly remembered that i tend to post too much...

Serge Yseron 01:01, 23 August 2011 (UTC)

Disguises

I think i just got a terrific idea for new gw disguises that will aslo do for the lack of mounts ! Please welcome the special aniversary epic mount. Yseron - 90.15.176.92 19:59, 25 August 2011 (UTC)

Humor

It work also in red. But i'm not sure the blue dots did appreciate...
Go on Franky, that's good, go on Franky that's good good good...

Serge Yseron 01:18, 30 August 2011 (UTC)

Fluxtility

All resistance is fluxtil.

Serge Yseron 01:25, 1 September 2011 (UTC)

What are you doing with a degen-heavy build when everyone has +200 health? Koda 21:27, 3 September 2011 (UTC)

Suprisingly enough, it worked fine. But to be honest, it shouldnt have. I guess it had more to do with the matchmaking throwing whatever teams it found however it could, provided that there is a matchmaking system. Yseron - 90.15.190.22 02:12, 4 September 2011 (UTC)

Bruce Willose

Brucy. Yseron - 109.213.154.66 17:25, 1 September 2011 (UTC)

AB

a game where there is nobody to play with is not interesting. Yseron - 90.15.59.103 19:31, 2 September 2011 (UTC)

Piss me off. Yseron - 90.15.59.64 13:44, 3 September 2011 (UTC)

2 key moments of the day passed and still nobody want to ab. If we must endure this every first week of the month, every month, i suggest that anet buy itself some designers that dont freak out ab players. No point doing one step foward with changing side at will, and 2 steps backward with a flux that apparently did empty the ab outposts. Yseron - 90.15.190.22 17:10, 3 September 2011 (UTC)

This is the weekend and i ususally play an average of 12-15 battles on saturday. Today: 2 battles. Time to find something else to play. Yseron - 90.15.190.22 18:21, 3 September 2011 (UTC)

Try the Fire Emblem games if you want some good challenges. The SNES games are by far the most challenging, but if you want to get a good learning curve, the GBA games are more interesting to start off with (stay away from FE8 though, it is trash). Koda 09:29, 4 September 2011 (UTC)

This will conclude this weekend, the privilegied moment of the week when we can usually ab a lot.

Serge Yseron 16:11, 4 September 2011 (UTC)

2D polygonal detection

While waiting for the ab outpost to populate i gathered some old code that was used in < NDA >, and made a sample code with some explanations, here. Yseron - 90.15.190.22 02:05, 4 September 2011 (UTC)

Read the docs guys

While toying around with an old psp dev-kit that was left abandonned at our studio, i came across an r&d project that was left unfinished. What stopped them was an un-explained hang in the libvfpu while processing a quaternion. After reading a doc it appears that these quaternions are 16bytes aligned structures, wich are not aligned any more if the class that contain them is not itself aligned. A simple aligned allocation did the trick. What surprised me next is how models display lists were constructed: the doc specify that when you call the sdk function that display a model, it in fact build a display list for that model. So instead of using cached lists and building only the part that is entity coordinates dependant they rebuild the entire display list each frame. I guess frame rate have never been the focus of r&d... Yseron - 90.15.190.22 02:58, 4 September 2011 (UTC)

I love you guys

We just have to wait for the luxons now....
Wait...
Wait again..
Waiiiiiit......

Wait.....
Waiiiiit.....
Waiiit.....
I dont know if anet must be proud or ashamed of having created the only game that you launch and then...
Nothing happen.

Serge Yseron 19:04, 5 September 2011 (UTC)

Fixing the fixs on fixed values

Some times ago, when open source content was still allowed for the developpment of titles on DS, i did get the occasion to toy with box2d as a contractor. I will remember two major issues with this one. Firstly, never assume: measure. We ended up gaining quite some fps after spoting several bottlenecks. Secondly, dont think that people will never dare to create a class that embed a float in order to support fixed/floating point values. They did..... This means that you can no longer always control when and how conversions will occur, wich mean that b2d spent 50% of its time to do conversions from float to fixed to float to fixed...etc. Here is a sample of the horror i went trhough before b2d begun to become a viable solution for the needs of the production. B2d can also be proud in that it almost made our technical director die from a heart attack, wich i may suceed to do one day so that i can write things like: const float echo_mending( sqrtf( powf( 2.0f, 2.0f ) ) );

I cant even download that file and it still scares me 24.130.140.36 20:18, 6 September 2011 (UTC)

Shadocks on the rock

Because it seems i get as much busy as i was 10 years ago, i decided to make me a framework that i will be able to reuse, fixing and cleaing up things in the process. I'm going to check if i can reiterate a version of a pathfinding that i made some years ago ( it wont use an AStar, but portals and cluster portals so, no, it wont be memory intensive and yes it will be fast ).

I uploaded a version of the framework in its current state here and moved it to an Ogre3D solution for the fast prototyping this middleware excel at, so that i can access it from work. Some screenshots should follow in the days to come.

Serge Yseron 03:48, 11 September 2011 (UTC)

I am almost finished setting up a boring framework (all files updated, including all kind of fixes), and i will soon be able to focus on what is really fun: the flocked / portalized / threaded path finding. I made sure the app does not crash at any given moment, and in particular when exiting it. I should be able to attack the pathfind job / clients requests sync tomorow or this weekend.

Serge Yseron 23:58, 13 September 2011 (UTC)

If somthing was buggy it was me. But it was so much fun to seek for the solution by myself than to rush on a doc like we do at work, and in the end, the solution that i found afterward are pretty close to mine. There is a redundant check on the zero shift that have been removed in the version that i uploaded. I now understand why my director was smiling when he got a glimpse at the early drafts of the sh64 functions. I think we will have a good laugh tomorow.

http://serge.billault.pagesperso-orange.fr/Quake3/sh64.txt

Serge Yseron 01:31, 15 September 2011 (UTC)

When examining the assembly code of msvc memcpy/memset/etc i saw there was an overhead for misaligned buffers. Since i use exclusively aligned buffers and pools, i will ship some versions of the mem* functions that dont need the head/tail overhead for misaligned buffers. They will be part of the <aligned> namespace so that i cant use the wrong version accidentally. I also see that there is a special support for intel platforms with buffers less than 256 bytes in size through MMX but i'm not sure yet of how i could take advantage of that on larger buffers. Yseron - 80.11.87.230 10:58, 16 September 2011 (UTC)

I have finished tracking the last bugs in the <aligned> namespace (ex: a GetPow2Alignement() used instead of a GetNearestPow2alignement() leading to a CObjectPool having a stride shift of zero in some circumstances ) and reuploaded the files. I also finished the instances auto registration and batch processing functions that allow instances to be deleted while being processed <- removed: this is a real bad idea. I know it has nothing to do with the project itself but the thing is that i am making a framwork at the same time, and i prefer to have clean base before going on with the rest. Since i can now see what's going on on my little grid, i can now begin the clients / pathfinding agent sync with a dummy path find function just to check that request are queued/processed/canceled/dequeued without any error. I got my time when at home, so i will use it to make things even more clean and reusable than when at work. Yseron - Serge Yseron 22:57, 16 September 2011 (UTC)

A yield in the agent job processing main function was missing. Creating jobs will now be easy for my futur toys. A base job got a lockable job request queue that is shared among one or more several agents (read: a variable number of threads performing tasks of the same nature). A verification on the thread handle of an agent led to hangs up in this agent when a job was not started. I had to redesign the various classes involved in order to get the more logical and clean architecture possible for jobs. Yseron - 90.29.51.24 14:01, 17 September 2011 (UTC)

The synchronisation between job and clients is finished and it was real fun seeing all the entities running after each other. I cleaned up the ogre framework and added the missing destroyScene(). I have uploaded the Ogre3D version that is fully fonctional so that if someone got the Ogre SDK installed, he will be able to launch it, and more importantly, to see why pathfing NEVER stall a game. The most interesting part are in Threads.h, and at a lesser extent, in CENtity.h where they submit requests and check for jobs completions (as opposed to a blocking "wait"). You can change the number of agents that share the same job requests queue but you rarely do because it means that if you got more requests processed simultaneously, it also mean that an individual request will take longer to complete. Now it's time to make a real pathfinding that use the grid (althought in the end it will only serve to create rooms separated by portals). Yseron - 90.29.51.24 18:30, 17 September 2011 (UTC)

I did not like the way a client could reset a request status so now job requests status is read only from a client point of view. I did move the seters into a protected section that is accessible only to jobs and job agents, wich also enforce a better consistency of the jobs status and allow to reset it automatically when appropriate and permitted. Yseron - 90.15.178.109 23:31, 17 September 2011 (UTC)

I have almost finished the auto generation of portals (yellow lines on the screen). We are only interested in junctions between rooms, because when inside a room, we know we can go onto a straight line in order to reach the next room, or in order to reach a target inside the room. Yseron - Serge Yseron 05:54, 18 September 2011 (UTC)

I dont have any more trouble with rooms (and hence portals) generation. I had to recall how i was implementing boolean operations on shapes in order to fix T junctions (that's 10 years ago...), but it's really just a matter of a few flags to set.

Yseron - Serge Yseron 20:59, 18 September 2011 (UTC)

The automatic generation of areas and portals is finished. I did upload the zip here. There was a memory leak in CreatePortals() where i did not delete bitsfields upon exiting the function, it have been fixed. Final and easiest step: use portals to navigate, or rather, to make job' agents build paths for clients that posted a request in the jobs queue. I dont know what will keep me busy after that. Maybe i will toy with the doom3 sdk and try to add a skill system so that i can life-syphon hell knights or kill npcs during cinematics (well in fact i already done that, that's how i ended up with the hero duplicate ( during cinematics they hide your character, and switch to another one ) runing with a grenade in the hand without his head while the npc was briefing me).

Yseron - Serge Yseron 20:18, 19 September 2011 (UTC)

95% finished: overall, 2 major issues - 1) an enter critical section performed way too late in the CLockable have been fixed. 2) largely not yet optimised and path portals list is to be reviewed because some entities somtimes jump a portal (wich inevitably lead to them walking across a wall). A test with 256 entities revealed no frame rate concern at all. Pretty much predictable. The CTArray not call constructor and desructors. It does it without noticeable overhead on PC but i wouldnt use this feature on slow toasters like portable consoles. I did update the files here

Serge Yseron 00:23, 21 September 2011 (UTC)

100%. I did upload the code here.
2 or 3 words about this basic pathfind:
- the concept of flocking (close entities with a distance threshold ask one of their own if another entity have the same target and got a path to it, then follow this entity) is absent, but you can implement it if you want by making RequestPathToPos() return the path of the entity that know the path to the target return this path and group it with this entity into a flock for inter entity avoidance along the path).
- paths that dont have a chance to be the shortest can be detected early, freeing the memory used by them as soon as possible before job completion add to the fact that this terrain representation already have fewer pts than a classic grid, and hence comsume less memory over time
- entities that have a read only access to the path finding request datas are usually started as soon as 2 or 3 pts are available (depend on entity speed ) so that the fact that they are queued and the associated latency dont show too much. When the final path is available, a roll back to the nearest pt of the shortest path is performed, that's why you sometimes see entities that look hesitative in some titles, wich add a human like behavior.
- unless you optimise it to the bones, make it multilevels/multigrid compliant, dont use it in any real production. Unless you make games where you can slash someone 100 feet above on a bridge...
Yseron - 90.14.228.232 17:46, 21 September 2011 (UTC)

Release version: - there was a dead lock in release only that i did get ride of.
- added flocking and an option to enable / disable it
- added start pos and end pos in both paths returned by the pf job and the dbg path because previously, only portal nodes were registered, wich was not a bug but a lazy move because this path finding thingy is just an exemple
- results: in release version, pathfinding alone without scene rendering 2775fps, 920fps with rendering as of 25/11/2011. But the fact is, you usually dont have 256 entities pathing at once.

It's there for whoever have some time to waste with it. I could have done a version that use only nodes but it would be like i'm giving free stuff to bots, and i dont like bots. A missing halt have been fixed, preventing exploration to go further when branching would always lead to a longer path, allocations dont yet use a reserved space to limit fragmentation and memory access time but i now reserve path datas with a single alloc by using the progression level that is incrememnted each time we go one branch further. An additional thread switch between each branch exploration ensure that each agent get a chance to begin a new job before the other agents finished their own job. In the "gameplay" code a minor bug was preventing entities from selecting the actual farthest entity as a target. The found paths registration have been extracted into another function for better readability. A bug related to how areas reference portals contained in a buffer that could be expanded before all portals could be referenced have been fixed. This exemple is now decent enough, i may tweak it again in the futur but i'm not sure. Serge Yseron 15:51, 24 September 2011 (UTC)

Parallelism

When making modern games is no more just about fancy rendering, and more about rich wolrd interactions trhough parallelism and intensive datas processing, or why we are more and more interested in threading and lock free data structures here... Yseron - 109.213.26.145 22:20, 23 September 2011 (UTC)

Macintosh

XCode on Mac is stunning. I dont know if it's only in appearence (read: is it still stunning when developping large scale applications), but i have yet to see its equivalent on PCs. Sometimes i think that to get the best of the 2 worlds, we should have Steve Jobs and Bill Gates in the same team... Yseron - 109.213.152.50 21:12, 24 September 2011 (UTC)

Space partitioning

Sometimes the best optimizations are the simpliest one. Here is one that i am sure i am not alone to oversight from time to time: in order to maintain some infos about wich game entity is in wich space partition we often check all partitions though diverses accelerators in order to see in wich one the game entity is after moving. Logic.

Now i ask you the simple question: Why. Should. I. Seek. In. Wich. Partition. The. Entity. Is. If. I'am. Not. Sure. The. Game. Entity. Left. Its. Current. Partition.

Yseron - 90.29.53.5 16:07, 25 September 2011 (UTC)

TLS defeating the non intrusive concept

It is possible to control allocations performed by thread local objects ( like a thread local pool ) that was declared with the __declspec( thread ) type specifier ( gcc equivalent: __thread ). Unfortunately in non managed mode, the only way to get sure that memory allocated by your thread local object will be freed is to get your object to allocate memory at the start of the thread proc, and to free it at the end of this same thread proc, or to passe it to the thread proc as a parameter. Now remember, we want something that is not so intrusive that the thread proc does not have to know about the thread local object declaration. Sometimes i like microsoft, sometimes not. Yseron - 90.29.52.198 00:36, 29 September 2011 (UTC)

WoC

I have some sympathy for the devs, they try really hard to keep us waiting with those new quests. - Serge Yseron 23:05, 30 September 2011 (UTC)

Dwayna vs Grenth goes jazz

here. Yseron - 109.213.153.25 15:59, 1 October 2011 (UTC)

lol liked the animation ^^ --The Holy Dragons 16:08, 1 October 2011 (UTC)

Nice movie, but death shouldve won that imo DemonicFahrir 16:16, 1 October 2011 (UTC)

Cutest roawr

in ages. Yseron - 109.213.153.25 19:43, 1 October 2011 (UTC)

Handles

There is a low memory footprint and speedy way to use handles in the specific case where you have, say, to make the AI or any other game engine part reference entities that can be deleted before various jobs completions. The one linked below did see some use in a real production under a slighlty different form but i didnt had really much time to implement it since it was a rescue mission for someone who had just learn about GOAP but had never thought that his entities could be deleted until 2 days before the submission date... header here and source there. By the way, i did fix something in the pathfinder, that was the occasion to add an auto maintained linked list into the framework that is coupled with a pool in order to replace an array that was providing faster results but at the expense of a readable and easely maintanable code. 90.14.228.32 21:43, 2 October 2011 (UTC)

Norns and christianism

There is something worse than not going to the valhala for a norn. The end is absolutely not what you think. Yseron - 90.14.228.32 22:07, 2 October 2011 (UTC)

Helping gw to go smoothly

If you are sure of what you're doing and if your anti virus allow it, you can filter the directories ( not recommanded if you use third party programes ) or the files for which he monitor activities, for exemple i told mine to ignore gw.dat. Otherwise, each time an operation is performed on it, the anti virus control what is done. It's sad to say but one of the reasons i was often laging as soon as i was getting damages from a lot of different sources in a short period of time was the dmgs hud ( the one that show the history of how many times you took dmg from wich source and spell ). I now rarely freeze when being the target of a spike since i deactivated it. Yseron - 109.213.152.18 23:48, 3 October 2011 (UTC)

Dumpbin

While toying with dumpbin on various exes including gw.exe i found this, wich led me to wonder what are the process involved, it could be the launcher, it could be something else, i dont know.

(...)
SHELL32.dll

               89837C Import Address Table

               9C11F8 Import Name Table

                    0 time date stamp

                    0 Index of first forwarder reference

(...)

                  4B SHGetMalloc

(...)

There is only 16 bytes of TLS but gw uses the full regiment of thread sync objects / methods. So you would think (unless those 16 bytes hold infos on a thread local heap/pools later created) that gw threads rely essentially on their local stacks with relatively small volumes of temp datas, locking shared resources otherwise. They got a clean exe whithout reloc section.

Yseron - 90.9.254.248 00:10, 5 October 2011 (UTC)

Sheeps and wolves

Sometimes you think of a word, and a lot of souvenirs comes back. For me it was the word "Predator". Flocking is in fact exceptionnaly easely optimisable with auto-maintained lists of... predators. Each time an entity predates another one, it is registered into the list of predators of its new target, and unregistered form the list of predators of its former target if it had one ( a neglectingly small overhead ). Flocking is in fact fast because now, you just have to cycle though the list of predators of your target in order to get an already existing path to it. I should have recall that from the start... Yseron - 90.15.189.79 20:55, 5 October 2011 (UTC)

Of flocking and latency -> the pitfall to avoid: if all the members of a mob request a path to the same target with the flocking option enabled, what's going to happen because of the pathfinding being processed into an agent thread, is that since the path finding job might not be completed for their predecessors, they might choose to submit their own path find request. This is a tremendous error because it completely negate the flocking bonus. Instead, if you detect that a colleague submitted a request for the same target, regardless or whether the related job is completed or not, you should not submit your own path request, but rather wait for your colleague's one to be completed. Further more, if your colleague' request takes a long time to complete, choosing to submit your own request would not speed up things since requests are queued and nothing guarantee that one of the n-other agents are not themselves busy with other requests. Yseron - 90.15.189.79 22:15, 5 October 2011 (UTC)

I had a misplaced bracket leading to nb enter critical section being different from nb leave critical section leading to dead locks if the same thread was locking an object more than once. I need to buy me a new pair of eyes. Yseron - 90.15.63.101 21:01, 6 October 2011 (UTC)

Of prototypes and things that actually work

I was going to keep me busy with something else when i finally noticed something odd. A small, lost and poor isolated entity went straight into a wall, defying the laws of physic. This would only occur with flocking enabled and was a bug very, really hard to catch because of its very nature tied to race conditions, not inside threads, but inside the entities management loop itself. Because i like to finish things and i like things that work i begun to track the phenomenon down with a battery of debuging patterns. The flocking is now flawless as i did a soak test of one hour with a large amount of asserts. The thing i wil remember if i ever have to code something similar one day is this: a leader can switch target, or cancel its target before it could dispatch it's path result to its fireteam, but the fireteam have to be able to continue with a shared reference counted copy of the leader's previous pathfinding result. It was tedious to the point that i almost decided to pass on, but the advantage is that now if i want to unleash 1024 enities on a target..... i can. Yseron - 109.213.26.183 12:43, 9 October 2011 (UTC)

Paragons in the mix

Crazy Fro' has gone from techno tyrolean paragon to master DJ Sogolon Mac Bass Bumper Max Energysing Final... Yseron - 109.213.154.198 20:51, 11 October 2011 (UTC)

2048

replaced some dull areas checks with a bitfield in the pfjob to boost paths calculations, fixed/enhanced fireteams, replaced the place holder for pathing states in Centity, cleaned up etc.

What i want next is to login in gw2, click on the alliance battles hot-join button and get something that gives me the kind of impression this music has on me.

Serge Yseron 20:47, 12 October 2011 (UTC)

Software bugs

Even if we know what he is talking about, we will continue to listen and listen again, until we humbly admit that bugs we made, bugs we make and bugs we will do. Yseron - 90.14.226.37 21:44, 12 October 2011 (UTC)

Activate the roger troutman

Things we all have been through. Blizzard is not far behind Arkane Studio with a npc that pathed into the ground. Yseron - 90.14.226.37 22:06, 12 October 2011 (UTC)

Tom Cruise has become a scientific argument

1rst class lol here. Yseron - 90.15.61.134 22:30, 13 October 2011 (UTC)

I think this one

will be more generic...

Serge Yseron 02:18, 16 October 2011 (UTC)