RBerenguel wrote:There are still network strain issues, even if security isn't a concern. Even if they are compressing the data, doing what is almost an X redirect through what I expect is a secured connection is a sh*tload of bandwidth just for playing a browser game.
I wonder if the Nova/OGS guys could just implement a GTP layer via http(s) so anyone could write a barebones client with a little patience.
I'm not sure how feasible that is at the moment since the site's features aren't stable enough to have a stable API. Perhaps in the future.
RBerenguel wrote:There are still network strain issues, even if security isn't a concern. Even if they are compressing the data, doing what is almost an X redirect through what I expect is a secured connection is a sh*tload of bandwidth just for playing a browser game.
Re bandwidth concerns, most of the data you receive from the site is cached so you only have to download when it updates. Once connected to a game we're using websockets to exchange data, so it's a very small amount of traffic, on par with native clients when you're playing live games. For correspondence play you're looking at having to load each game up in turn, so this takes a bit more bandwidth - about 15KB per page, so playing 4 moves is about the same amount of traffic as loading up a random image of a cat on the internet.
RBerenguel wrote:There are still network strain issues, even if security isn't a concern. Even if they are compressing the data, doing what is almost an X redirect through what I expect is a secured connection is a sh*tload of bandwidth just for playing a browser game.
Re bandwidth concerns, most of the data you receive from the site is cached so you only have to download when it updates. Once connected to a game we're using websockets to exchange data, so it's a very small amount of traffic, on par with native clients when you're playing live games. For correspondence play you're looking at having to load each game up in turn, so this takes a bit more bandwidth - about 15KB per page, so playing 4 moves is about the same amount of traffic as loading up a random image of a cat on the internet.
He's referring to the serverside browser mentioned previously, not to nova itself.
skydyr wrote:He's referring to the serverside browser mentioned previously, not to nova itself.
Exactly. I guess the amount of data once the game is loaded is negligible (I assume just the data to manage the game area so close to 0 with today's speeds), but a remote browser has to send at the very least 1 image per change, not sure how optimised the other side may be (it can be 30 frames/second or 1 frame for every n-pixels changed in the virtual screen.)
@skydyr, the API can be hooked a a middle layer GTP API. So, a generic GTP API is implemented and more or less fixed to the outside. When the inner Nova API changes, the GTP API changes how it connects to the Nova API, the GTP API doesn't change to the exterior. It's not like Nova can't do currently what GTP asks for.
Geek of all trades, master of none: the motto for my blog mostlymaths.net
The map / graph for the open challenges is a neat feature in my opinion and a way to integrate slow and fast play is very much necessary when the time limits start with seconds and goes up to days. Imagine having only a list, which isn't necessarily sorted by time limit, where dozens of correspondence game requests make it hard to select a live game challenge.
There are features from OGS, that I miss on "nova OGS", however. Reviews are owned, which is probably a good way for teaching but the typical collective review featured on OGS is just not possible. Also, old reviews from OGS seem to be lost, unlike the old game records. Other than that, it already is working quite well.
tapir wrote:There are features from OGS, that I miss on "nova OGS", however. Reviews are owned, which is probably a good way for teaching but the typical collective review featured on OGS is just not possible. Also, old reviews from OGS seem to be lost, unlike the old game records. Other than that, it already is working quite well.
I haven't checked myself, but I read somewhere that old reviews do not show up as "Review" but the comments are actually inside the game itself, so they are not really lost.
tapir wrote:There are features from OGS, that I miss on "nova OGS", however. Reviews are owned, which is probably a good way for teaching but the typical collective review featured on OGS is just not possible. Also, old reviews from OGS seem to be lost, unlike the old game records. Other than that, it already is working quite well.
I haven't checked myself, but I read somewhere that old reviews do not show up as "Review" but the comments are actually inside the game itself, so they are not really lost.
skydyr wrote:He's referring to the serverside browser mentioned previously, not to nova itself.
Exactly. I guess the amount of data once the game is loaded is negligible (I assume just the data to manage the game area so close to 0 with today's speeds), but a remote browser has to send at the very least 1 image per change, not sure how optimised the other side may be (it can be 30 frames/second or 1 frame for every n-pixels changed in the virtual screen.)
@skydyr, the API can be hooked a a middle layer GTP API. So, a generic GTP API is implemented and more or less fixed to the outside. When the inner Nova API changes, the GTP API changes how it connects to the Nova API, the GTP API doesn't change to the exterior. It's not like Nova can't do currently what GTP asks for.
Ahh, yeah that's kind of wild, and agreed, seems like a whole lot of bandwidth for something like this I didn't even know such things existed.. sorry for the confusion
skydyr wrote:He's referring to the serverside browser mentioned previously, not to nova itself.
Exactly. I guess the amount of data once the game is loaded is negligible (I assume just the data to manage the game area so close to 0 with today's speeds), but a remote browser has to send at the very least 1 image per change, not sure how optimised the other side may be (it can be 30 frames/second or 1 frame for every n-pixels changed in the virtual screen.)
@skydyr, the API can be hooked a a middle layer GTP API. So, a generic GTP API is implemented and more or less fixed to the outside. When the inner Nova API changes, the GTP API changes how it connects to the Nova API, the GTP API doesn't change to the exterior. It's not like Nova can't do currently what GTP asks for.
Ahh, yeah that's kind of wild, and agreed, seems like a whole lot of bandwidth for something like this I didn't even know such things existed.. sorry for the confusion
Don't worry, no problem. But think about GTP
Geek of all trades, master of none: the motto for my blog mostlymaths.net
RBerenguel wrote:I wonder if the Nova/OGS guys could just implement a GTP layer via http(s) so anyone could write a barebones client with a little patience.
GTP is not suitable for writing a client, AFAIK. It is meant to connect go engines only and has no provisions for such things as kibitz, chat, observing games, etc.
RBerenguel wrote:I wonder if the Nova/OGS guys could just implement a GTP layer via http(s) so anyone could write a barebones client with a little patience.
GTP is not suitable for writing a client, AFAIK. It is meant to connect go engines only and has no provisions for such things as kibitz, chat, observing games, etc.
Would allow playing in any case. Good enough I think...
Geek of all trades, master of none: the motto for my blog mostlymaths.net
RBerenguel wrote:I wonder if the Nova/OGS guys could just implement a GTP layer via http(s) so anyone could write a barebones client with a little patience.
GTP is not suitable for writing a client, AFAIK. It is meant to connect go engines only and has no provisions for such things as kibitz, chat, observing games, etc.
Would allow playing in any case. Good enough I think...
Also, no way to get a listing of open games, no way to join an open game, no way to negotiate time settings, no way to refuse an opponent.
With GTP, you can only act exactly like a bot: create an open game at your desired time settings, accept any opponent, never chat.
The other big thing, for anyone having known OGS, is the current state of the tournament tables. Lists are not a good substitute for tables, much less compressed and harder to parse than an actual tournament table, there are some advantages of seeing all rounds at once, but you don't see how the tournament is going anymore. Tournament play was after all the main advantage of OGS over any other server I know. So, I guess this will be one of the issues that needs to be addressed to retain the OGS player base.
I didn't even know such things existed.. sorry for the confusion 
