cabos and acqlite: using LW core modified by heavy_baby
 

Salut all, Philippe Verdy, one of the GDFers and responsible of LW translations, told at the GDF that Cabos was using a modified LW core that could hurt the network: http://groups.yahoo.com/group/the_gdf/message/21828

Since today, heavy_baby also use the name "acqlite" a software abandonned by Pozytron, its former programmer, to gain marketshare. http://macupdate.com/info.php/id/12747

I got him to respond to some of Phillipe's claims:
http://macupdate.com/reviews.php?id=12747&pid=117304


----
heavy_baby:
>>Also it performs automatic download requeries every hour, instead of waiting
for user interaction when a download fails.

It's not true. acqlite does not automatic requery.


>> It also silently ignores minimum quality and minimum speed from results

This should be fixed in the next release.


>>It also increases the number of connections with matching prefered locales
from 2 to 4

It's not true. acqlite prefer default value '2'.


>>It also shares the content of the "Incomplete" directory as normal files

Resistered incomplete files will not be shared. Even on download folder.


>> removed the support for Chat
LimeWire Chat is not needed, and LimeWire user can disable it on the preferences.


>>WhatsNew searches,
What's New search will be added in the future.


>>Daap sharing for iTunes..
Daap is already supported.

------
me:

Are you trollling? You avoided to talk about the worst tweaking Philippe was talking about:

"uses an
unusual default for the number of connections that leaf nodes maintain to
ultrapeers: 5 instead of 3."

That's sucks. Just as I told Dave of Acquisition, but he didn't listen completely because he was afraid of its users reactions (he dropped from 6 to 5 and first told me he would lower to 3 -- has yet to happen).

"It also computes the dynamic queries from
ultrapeers using 300 expected results per connection instead of 150"

That's what I call network congestionning, this can cause way more than 2 times the traffic to get 2 times the results.

"When downloading, it swarms with 2 additional sources for modem or T1 or T3
users."

Thus lowering download slots for other users....

"More seriously, its ultrapeer will forward all query hits it receives, even
if they don't match a past search (using GUID matching)."

Bad.

"It also shares the content of the "Incomplete" directory as normal files
(simply because it locates this directory within the shared default download
folder, and not within its parent): are users happy with those results for
incomplete files?"

hum...

You should respond at the gdf ASAP, you would have done so long ago if you were actually clean.

Hope you "get it" unlike Dave Watanabe. You can always talk to me about it at gnutellaforums.com if you feel so or don't understand the consequences of the changes you made.

-------

What do you think? :p

> It also shares the content of the "Incomplete" directory as normal files

No. It's not true. You can see FileManager.java:

if (directory.equals(SharingSettings.INCOMPLETE_DIREC TORY.getValue()))
return Collections.EMPTY_LIST;

This means incomplete folder is *not* shared.

> "It also computes the dynamic queries from ultrapeers using 300 expected results per connection instead of 150"

It's not true. You should check acqlite source code of QueryHandler.java.

But some issues you have mentioned should be fixed in the future.

Please tell Philippe at the GDF. I normally do a testing of other Gnet apps, but I don't have time (nor I don't know where to look at the source code) to test Cabos and Acqlite settings right now. Only thing that I can say is that they connect to 5 UPs per default. Looking forwards for the "fixes" :p

ciao

> I don't know where to look at the source code
If you are sure, I can send you the patch diff of acqlite core code. To save your time.

>Only thing that I can say is that they connect to 5 UPs per default. Looking forwards for the "fixes"

>When searching, it sends What's New requests to 4 ultrapeers instead of 2,
No, it's not true. acqlite prefers default '2' peer What is new query.

> "uses an
unusual default for the number of connections that leaf nodes maintain to ultrapeers: 5 instead of 3." That's sucks.

It's not sucks.
I know 5 connection is LimeWire's default. I think it's not problem.

LW default is 3. Bearshare is 2. Now you probably talk about Pro which is 5, but they are less than 10% of Limewires and help to pay the programmers for improving the code YOU use.

Why are you avoiding talking about Cabos, the changes aren't the same as Acqlite?

Ciao

PS: if you could send me the diff here or by PM, that would be great ;)

PM does not accept 164KB patch file.
Please send me your email address.

Sorry, I'm saying about network, not money issues.

If 5 connection is not good for network, why limewire pro & acquisition do this?
3 is maximum safe value, isn't it?

The network current topology was designed so every user fas 3 UP connections as a leaf. It allows good network reach on searches and lighten the UPs bandwidth while avoiding users to get DDOS by requests for some files. Just compile LW for yourself as I do, you'll get 3 UPs without GUI options to change it. The 5 UP connections on Pro version is just a "Thank you for supporting LW development". In fact, when a user buys Pro, it also help you since you use the core developped by those paid programmers.

Acquisition does it only because Dave is a moron, I could tell you so many bad conversations I and others had with him :rolleyes:

If you register here, I'll email you my email address heavy_baby (one of the privilegdes to be a supermod, héhé). For security reasons, we never post email addresses here.

Ciao

I have already registered and added your name to buddy list.

Thanks,