The users complaining about lag have been asked to provide tracert results. They have not done so. If they are unwilling to go through simple troubleshooting steps to try and diagnose the problem, there is not much help to give.

From what I can understand, they're unlikely to benefit from seeing where the connection problems exactly came from (4th hop vs. server). I don't know about others, but I sure as hell won't change my ISP provider for just KGS, especially when I don't know if that will fix the problem/create problems on other sites.

The request seems to be some sort of lag-handling protocol, because frankly lag is really not possible to eliminate.

Hello,



I'm sorry to say that you've misunderstood my statement. KGS cannot "solve" lag. It is an inherent consequence of having shared connections on the internet.

Companies which need to minimise lag (e.g., hospitals, airport flight management, missile defence) pay really large amounts of money for specialised equipment and experts to maintain it. You, on the other hand, presumably use consumer computer equipment, a public ISP and KGS runs on a commercial server. In these circumstances, lag spikes will always take place. Although you state that other go servers don't have lag, no supporting evidence has been presented to support these statements and I find the statements' validity unlikely.

One of the difficulties in resolving technical issues is that sometimes people start the analysis with a pre-determined solution and therefore spend their time seeking evidence that will support it. I'm afraid that it appears to be the case here. I understand that you genuinely believe that lag only exists on KGS and can be stopped if only the developer wanted to. I suggest that you take a moment to reset your assumption and view the thread again. A number of people have tried to explain why this assumption is incorrect.

If I can make a suggestion, if you find all these comments about trace route and hops and routers confusing (and I certainly don't blame you for feeling overwhelmed), don't say that you're very knowledgeable about networking. By claiming an expertise, people understand that they can use more complex terminology with you. By avoiding overstating your knowledge, then your chances of encountering jargon are smaller and people generally stop when you say that you don't understand.

Hello,



When I was younger, a character on a television series used to say "make it so" when he wanted something to take place. At that point, it was assumed that everything required would just happen. Presumably, somewhere in the background, lots of people were working really hard and spending lots of time and money until -presto- the problem was solved. On the show, of course, it's simply a screenwriter's trick to solve a plot issue quickly.

In this instance, several people have sincerely expressed their wish that a problem which affects them be solved. They recognise their circumstances don't affect the vast majority of people on KGS, but it ruins their fun on KGS. It's frustrating to have an unreliable connection and everyone in this thread presumably sympathises with the people who are complaining. I certainly do. Unfortunately, this problem is quite complex (as a number of people have tried to explain).

The "unhelpful" responses you point to are simply people trying to provide practical advice on how to work within the current environment. Everyone would honestly rejoice if solving lag issues was simply a matter of changing a 1 to a 0 somewhere in the code, but it's more complicated and time-consuming than that. Therefore, people have been trying to provide what assistance they can. I agree with you that this advice doesn't actually solve the problem, but people are doing what they can to help.

I like to squeeze in a quick game during lunch break at work so I certainly would not appreciate lag compensation if being stuck with someone playing over a satellite.

If my opponent have a constant lag of 10s (no joke, I have seem it with satellite connection) that would make the game twice as long.

Some of us have real time constraints.

Here you come again insulting my intelligence. You seem very proud of your networking knowledge since you've learned about tracert, so listen carefully.

Tracert is unreliable when checking client/server connections:

-The road the packets follow is asymmetric

-Tracert only shows the route MY packets take to go to KGS, not the return trip

-The road taken one day can be completely different the other day, even if I'm very near the server

-Routers emit ICMP packets passing through the interface used to link to me. It's not necessarily the same interface I use to send normal data.

-Routers don't handle ICMP packets the same way they handle packets of data. If my network uses Quality of Service (QoS) to facilitate transfer of data, the result from tracert (ICMP) will not reflect my connection quality.

-The creation of the ICMP packet (TTL exceeded) is a complex process that uses CPU resource from the router. (when normal traffic uses specialized hardware) There can be some additional delay that wouldn't happen with normal traffic.

-If the IP address of the ICMP TTL exceed is private (http://tools.ietf.org/html/rfc1918) it can be blocked or unidentifiable.

-Some routers don't answer or just limit ICMP packets. (a "*" appears in tracert result)
Those *'s don't necessarily show that there's something wrong with my connection.

Do you understand why asking users to use tracert to find the cause of KGS lag is ridiculous?

I'm impatient to see what other nonsensical arguments people will try to feed us.

Conclusion:

NO I will not waste time using traceroute to see what route some random ICMP packets take when normal traffic most likely takes another route and I will not give the results to people who don't even understand basic networking.

NO my connection or hardware is not at fault. I know my setup better than you do.

NO, the normal lag on other servers (hidden by clever software implementation) has never been the cause of me losing games or being unable to contact the server at all. I've never had any connection problems on other servers than KGS.

YES, KGS needs some DB and client/server optimization to eliminate the abnormal lag many people suffer from.

This marks the end of my participation in this ridiculous thread.

Lol, this thread rules!

*chortles*


I'll get the popcorn.

How the thread should have gone:
Post 1) Some people have legitimate lag-related issues (either the lag itself or the handling of game-play issues related to inevitable lag).
Post 2) KGS knows it, but time/money are limited so other priorities are being handled first at the moment. "Thanks for your patience."
Post 3) Okay. I still like other parts of KGS. I hope you guys get around to this soon. Thread over. The end. Boring and short.

Then again:

That about sums up the novel, but many people still enjoyed the remaining 700 pages.

This thread was very interessing for me.

I noticed the location, like me in Asia from the complainers and I have the same problem as them. Some very high Dan players from Asia regularly suffer from lag problems too in their games.

I would add that kgs policy on escaping is so rude for all of us that it is because we love so many other sides of kgs that we still try to come and play here.

good news is that wms is rewriting a HTML code, I hope that he will consider some new rules for the escape policy too as to protect and not exterminate players who have lag problems.

Thanks for all what i could read here.

That is the point of the exercise. if traceroute shows both routes, they wouldn't need you to do it.
They can only traceroute the path from the server to you.

And even if the admins are not trace routing you, there is still a 50% chance that the lag is along that path.



True.


Perhaps relevant when Leonard Bosack and Sandy Lerner built their first routers, not for today's routers.


I don't see how this applies to your case.



Traceroute is not certainly a perfect tool but that does not negate its usefulness in basic troubleshooting.

@BigDoug and others who are trying to help:

I think my "absolutely unhelpful" was too harsh, and I understand now that the attempts to troubleshoot is meant to help. Apologies for my wording choice.

Nevertheless, I think the attempt to individually troubleshoot is going to be a drop in the bucket unless it's true that the lag is extremely uncommon. Lemmata's comment feels pretty spot-on, but I personally don't think that high of the typical PR response - at least here the response feels genuine rather than a response without consideration.

Can someone explain to me the differences between a lag compensator (which do not help if congestion is so bad that connection is dropped) and using longer time or the Canadian Time system ?

If you have constant lag, use longer time.
If you have sporadic lag, use Canadian Time.

I think that the difference is not necessarily what happens conditional on finding a game. The difference maybe that it may take forever to find a willing opponent if a game with longer time or Canadian time is posted. It takes a long time for people to accept posted games as it is now. Sometimes it even takes forever to get an automatch game!

Of course, if you're someone who everyone wants to play like twoeye, yeah, that kind of solution will work like a charm.

I would like to add, as the OP who started this thread in a burst of frustration over a year ago, that while troubleshooting has been suggested several times, I have not gone this route simply because it's not high on my priority list. My assumption is that attempting to fix an intermittent problem such as this would involve more time, energy and possibly money than I wish to spend. In this regard, my position is similar to KGS's as expressed by BigDoug.

It is however important to point out once more that the problem is not the existence of lag, which can be reduced but not eliminated, but rather that lag has undesirable consequences for both the user and KGS. For affected users, the consequences are frustration, anger and unnecessarily lost games; for KGS, the problem is a poorer user experience. There are many steps both technical and practical that could alleviate these difficulties, and while some have been taken, others have not.

It may be true that KGS is not responsible for lag, but it is not the case, as several posters starting with Mivo (who IMO most accurately hit the thread's nail on the head) have pointed out, that KGS could do nothing to reduce the effects of lag. It's just that they, like me, don't want to.

Placing blame is unproductive, but I agree, it has been entertaining.

_________________
Patience, grasshopper.

I'd been more or less avoiding this thread because it seemed a bit like a rant thread (and hey, who doesn't love a good rant thread?) but just one thing to consider on these lines (if we're aiming for discussion). A proper lag compensation system requires access to the system clock, which pretty much by definition means native code (which in turn means no longer having a portable java application...or if you use the applet it isn't even an option at all). I am familiar with other servers for other games that do compensate for this, but I don't know of any that are cross-platform compatible (i.e. they are all windows only from my experience as a windows user), and they require downloading / installing an executable program (instead of a of just running a java webstart application).

Can you elaborate on what you mean by "system clock"? Obviously java has the getTime() method, but it seems you might be referring to something more low level, and presumably less hack-proof?

It even made me change my signature.

The reason that users experience lag on KGS more than on other servers is the basic design principle that KGS will only display your move after one roundtrip to the server and back. (As pointed out somewhere above.) If moves would be displayed directly in the client, things would behave more smoothly. Exact timing of moves within the last byo-yomi second would be no problem at all (try out the asian servers, like Tygem, to see what I mean).

I would say that this is one of the things which made a lot of sense in the past, when connections where much more unreliable and slower than they are today. I guess that if wms would design a server today, he would display moves directly in the client.

_________________
Stay out of my territory! (W. White, aka Heisenberg)

This does not reduce lag, only the perception of it. Of course better feedback gives a faster and smoother experience, but also a more shocking one when it fails. You might see the move on your board, and then lost on time. That would be incredibly upsetting.

_________________
Founder of Kaya.gs

The client knows how much time you have for your move and can control that exactly - why would you lose on time?
Lag would be noticed by how your opponents clock runs slower / adjustments to your opponents clock.

Have you played on Tygem? Did you ever lose on time?

_________________
Stay out of my territory! (W. White, aka Heisenberg)