Support says: "You can download a fairly simple I2P client called i2pd.  It's website is https://i2pd.website/.   We don't have anything to do with it's development, so do your own due diligence before installing it.  The original I2P client (Java-based) is available from https://geti2p.net/en/.   There are also other clients available, easily findable using any search engine."

So, calling all geeks who are familiar with I2P.

Is there a particular one recommended from your experience?  And Why?

What are the pros and cons of the ones commonly available?

Is anyone particularly recommended for use with Tixati?

I have used both Java I2P and I2Pd. I generally recommend Java I2P if you aren't sure. Java I2P is more user friendly and has more features in its webUI. I2Pd has a very minimal webUI and requires editing a config file. The main benefit of I2Pd is that it is written in C++ and has lower CPU/RAM usage than Java I2P.

From my experience, i2pd is faster, consume less resources.
It also seems that Tixati somehow works better with i2pd.
Yes, you have initial pain to setup i2pd, as everything is configured in conf file.
However, despite extensive web console i2p still requires fair understanding what yuou are doing in that console.
Either way, you have to spend some time reading docs. So the difference is really minimal here.
I've started with i2p and then switched to i2pd, because my apps written in c++ and I wanted to minimize overall setup into one machine without dealing with Java complexities and endless hickups.
What actually appeared, that Tixati does not use i2p very much. I see very rarely any connections.
If there is ip4/6 available, Tixaty uses those almost exclusevely.

I have talked to I2P developer and he says there is no difference in UP/DOWN speed of I2P torrenting when using Java I2P, I2Pd, or I2Pplus.

All have same speeds, if all have same configuration.

Only difference is, I2Pd is bare minimum daemon, whereas Java I2P & I2Pplus comes with whole suite of embedded applications such as email service, web-hosting and other things.

In any case you will need to read the docs carefully. I use i2pd only. Java I2P wastes 50% of its CPU cycles, annoying for 24/7 use and low end machines. Devs really do not focus on optimizing. Lots of the code is around two decades old and specifically targets 32-bit single processor machines.

Untrue. There are many delays in Java I2P(plus) that add up over time, hard to measure with the constantly varying transfer speeds.

- Java I2p is the reference implementation, not primarily targeting production with many code sections taken from academic research, written for clarity, not speed.
- The high number of threads combined with the much higher CPU usage causes significantly higher latency as each packet takes longer to travel across the router
- There is a forced 25ms pause in TCP processing every second (targeting Windows 95), that may be significantly expanded depending on the powersaving features of your OS.
- If interrupts caused by high network traffic hit a worker thread holding a crucial lock, multiple worker threads are stopped. Can take more than 10 ms for the JVM and the OS scheduler to get everything going again.
- There is no separate disk writing thread as mandated by good programming standards, so you can grind multiple router functions down to a halt by simply starting another process in parallel with heavy disk writing activity.

I think you should cancel your neural network entertainer service subscription immediately, and donate that money to nearest charity to make the world a little better place. Now you are actively making it crappier by copying and pasting nonsense. You're worse than a spammer.

If you don't understand that Java programs don't “target” the internal inconsistencies of specific systems because they are wrapped and abstracted by JVM, that 25 ms means hundreds of years for network related code, that Windows 95 had been totally irrelevant at the time I2P project started (if you had any reason to use an old 9x system, it would be 98), that most I2P members have not actually come from academic circles, and only reference the research, don't make yourself look like a fool.

I am going to see everything you post as LLM garbage until you directly state otherwise.

As for the main question, the answer is obvious. If you are a new user, the only way to understand what happens is to use Java I2P, see how it works, and read descriptions and documentation. i2pd provides minimal stats for people who already now what they mean, and how to interpret them.

You can run both nodes if you change the clashing port numbers. The easiest way is probably starting i2pd alone, shutting it down, editing i2pd.conf to change local services ports (web interface, HTTP/SOCKS proxies, control protocols). Public network port can remain random (unless you need it to be static to forward through NAT or firewall manually). Other option is to use the parameter to generate default i2pd.conf before the first run, and place it in the appropriate directory used for settings and data in advance. You can also use the parameter to set any non-default directory as profile location (and run multiple i2pd nodes with different settings).

The network seems to slowly recover from the attack after today's release, so stats might still be all over the place.