World: r3wp

ah, dot trick worked, thanks ...

R: I would add to our list of mail-list users, all altme users whenever 
safeworlds single point of failure archetecure rears it's head

(such as the last half day)

ah, finally back? What happened?

uf, pressed Bold button mistakenly ....

when all worlds are unavailable from everywhere it is because safeworlds 
is having network/hardware/software problems  on the machine that 
looks the current address of a world

Yes the Safeworlds nameserver application was done. The application 
runs on the same server than their website.

even if the world host address is working just fine in the next room

no, not apparently, as their website worked here, while Atlme did 
not?

Because the nameserver app was done. The application runs on port 
5500. Webserver in port 80 both hosted on the same ip-address 66.328.43.215

website need not be on the same machine seerving world addresses

nameserver app failed

I wonder, if Altme a) uses at least TWO machines to connect to for 
lookup purposes like DNS does b) can remember IP address so next 
time look-up server is not available, it can try direct IP connection 
.... if not, it is imo plain stupid as at least b) imo would not 
take more than few hours for programmer to add .... now Peter will 
come once again, and call it - "unreliable" :-)

From my observations of pkt trace.
a) No it doesn't
b) No caching either
Pekr "unreliable" assestion correct.

safeworlds did have a good long run with no downtime, but stuff happens 
eventually

not "pekr unreliable" - it was joke. It was Peter Wood imo, who called 
it unreliable and once Rebol2 failed, someone joked that now Peter 
Wood will come and call it unreliable, so I just wait when he appears 
:-)

Pekr- it is more a policy / bussiness pratice than praciical programming 
problem

there are some advantages to safeworlds, theu can provide/ force 
upgrades to their software

why business policy? Do you think SafeWorlds wants to keep it under 
control? Well, bad service then? I don't believe they don't have 
at least two machines. Alme client could contain list of alternative 
look-up server - it would cover most outages imo ...

and it keeps them in the loop, not obsoleating yourself may make 
good bussiness sense (I am useless at bussiness so I am geussing 
here)

I think the should build some redundancy in the name resolution part 
of the AltMe service otherwise there will be service cut-offs.

I think their feeling is that if they can provide reliable world-name 
service then it is the best choice.

and they think they can

Tom, there other ways to address the "obsoleating" policy that doesn't 
require service cuts.

however realithy does bite

I am not the one that needs convincing ;)

Tomc - if there is any business practice, I just don't buy it and 
SafeWorlds wants to stay in control. Then they should not promote 
product as private. IIRC, folks are reporting even their private 
running worlds don't run because of look-up server failure ...

but they are private in that the data exchanged after the world-name-server 
lookup does not go anywhere near the world-name-server... it is just 
words

I'd be really embarrased right about now if I had promoted Altme 
as a corporate communications tool.

Well... Partially truth. The name server is consulted all the time. 
At least from my pkt trace. I am looking more deeply into it.

If safeworlds were able to provide the service they are attempting 
to then we would not care for the most part

Well, imagine running some cooperative group of ppl and lose ability 
to work because of AltME outage. What is even more strange is, that 
SafeWorlds did not bother to post any kind of message on their website. 
RT did ....

effectivly true... if you have a connection and the world_name-server 
goes away you do not loose the connection you have.

yes, that is probably true .....

definatly

The msg are sent encrypted between server and client. But the name 
server is consulted all the time to know if there are msg waiting 
or something similar. I ma not quite sure why the name server keeps 
engaged after initial resolving.

It does? Strange ....

If you had a pkt tracer check out the traffic going to 66.218.43.215 
port 5500

coule be keep-alive or statistics gathering

quality control not needed for operation

yes, I have Ethereal, never checked on AltME yet ...

;Interesting every second the client send a request that looks like
cmd: live data:  [world-name binary64]
;and the server replies with this
cmd: lookup data: ok

In the case above the "client" is my private AltMe world server. 
So the traffic is being sent from my server to the nameserver every 
second. I guess is a way to keep the ownership of the free AltMe 
world names.

Enough hacking back to real work...

.

Tomc wrote "safeworlds did have a good long run with no downtime, 
but stuff happens eventually"


Thank you, and yes we did.  The error that happened today was fixed, 
and an additional check is now being made to prevent that from happening 
again in the future.


One thing I can clearly say is that each problem has been unique, 
and each time one happens (a severe failure) we stop everything and 
jump on it, and in addition to correcting the current problem, write 
code to prevent the same problem from happening in the future.


We have been down less than Live Journal, which has millions of people, 
and hundreds of servers for the same time period.  So we rock!

and you're modest too ;-)

Modesty is a disease, and it wastes time, and truncates communication, 
which are both ethical crimes.

heh.

Anyone here know anything about Kylix?