Hello

I would like to request hard limit on trading, as percentage limits does not work properly, maybe its just not imperative enough.

On same note, return of hard limit for auto bandwidth would be welcomed as well, as gate detection is broken (maybe just me, but usually like 300kbs above my max speeds)

Thank you

Confused.
Hard limit on "trading" when downloading is accomplished by you setting the upload limit.
Trading allows you to "prioritize" sending up to XX% to peers you are currently downloading from (trade) versus straight seeding to peers. NOTE this is ignored when using Seeding+ (constant seed no slots)
Default value of 75% trade means that UPLOAD_LIMIT *.75 will be provided for your downloads.
If you think that this means you WILL ALWAYS upload that ratio you cannot control other peers. You can increase likelihood by increasing priority of your downloads to see.

Many times too many connections cause too much churn with data. I would stick to the UPLOAD_LIMIT/TRANSFERS/connections?=20 KiBps (1 Mbps speedtest up is 128 KiBps) This ensures at max connection limit you are still putting out 20 KiBps per any peer connected to you. Which will complete a whole piece from you in less than 1 minute for 1 MiB pieces. Scale up/down accordingly for piecesize... Having seen 16 MiB pieces in the wild this can be quite a task. As the only time you get verification of data is once you complete a piece and transmit that information through the swarm.

I know trading can be limited to % of you upload limit, but it does not work properly, I can limit it to 10% and it will take whatever it want, based on some algorithm, who thinks seeding does not want it so trading takes it (and if i limit upload of downloading transfers, seeding immediately takes it...), its soft limit, and I want hard one, meaning for example, I have 400kbs upload limit, and I want 100 for trading never more than that, even if theres 200kbs not used by seeding, as there never is anyway, in my case

as for auto-limiting upload, it used to have hard limit, than it was changed to gate limit, but sadly another example of algorithm overthinking stuff, its never accurate, at least not in my case
i know my upload goes from 800kbs to something like 400kbs based on line usage of other users, so i use auto limit based on ping to keep connection fluid for me, and i limit download based on that, as i know my limits i can use ratio to set up reasonable download limit based on current upload, but when gate limit goes stupid, setting lets say upload limit to 1.5mbit, my download limit its also above my maximum speed and soon as i start downloading something everything starts throttling down until it hits my real limits, so would be much easier for me to set my real limits and never go above, and do not experience unnecessary throttling when it goes from gate limit to my real limit

universal automation is not bad, but setting things manually should be possible

Thank you for the explanation I was not understanding. I see the bytes-for-a-piece-in-timeframe was unneeded.

I think Since you CAN set a hard cap on your upload it should be a little more obvious the breakdown when separating TRADE v SEED v other (channels dht etc). ++
So another option here with set understanding for the limit and what it means would be helpful. My caveat for these types of things is that it's not always obvious where/what all the settings are for.
ex: see the recent revert of changed logic for tracker announce. It's an advanced setting now you need to setup for each/any tracker you want to contact. Before it was an all/none proposition.

Can you please try this for me: Right click on a seeding transfer go to bandwidth - allocation - upload priority and change it to anything above NORMAL... if you already use priority choose a level higher than the MAX you already use.
As I have read here and checked some myself. When you have a higher bandwidth priority it is a fibonnaci like curve for bandwidth allocation among transfers all other things being equal.

To check this type of stuff you would need to try on a peer/leecher heavy swarm where you already have the data to test. Spend ~5 mins comparing behaviour what you see NOW with what that one change does. Even going up 1 level usually ensures a 3:1 seed/trade ratio. Also please reply with more info if I still don't get it. :)

for autolimit:
ok i missed that you can set limit outside of auto limiter and it will work as limit even for auto limit rules ... my bad, does not fix bad gate detection, but works for me :)

for trading limit:
i use priority all the time, and seeds are always higher priority than downloads, but as i see it, main problem of trading/seeding limit is number of peers

example:

i have 20 seeding transfers, with 1-2 connected peers each, easily use all of my upload speed

then i start 1 downloading torrent with 200 connected, mostly trading peers

this one downloading transfer will take all upload bandwidth no matter the % limit for trading or priority, probably because 200 peers with low priority is still more than 20 peers with high

i usually just limit upload of downloading torrents manually, but working trading/seeding ratio would be much better

Outgoing Seeding/Trading % works OK for me. I use the default setting: Trading 75%, with reasonable total incoming/outgoing limits. I observed two curious things:
1. peers unchoked for charity (with blue LC status) count towards seeding % even when they are in downloading transfer, in other words there are trading peers and seeding peers, transfer downloading/seeding status isn't important (constant seeding also doesn't matter)
2. Bandwidth Priorities do alter Seeding/Trading %

I think peers number isn't important. I sometimes observed that one seeding peer got all 25% reserved for seeding while each of several trading peers got less than that. I also saw other funny situations, for example: only 1 downloading transfer yet there is 25% seeding bandwidth on the chart. This was caused by peers unchoked for charity, even funnier each of those got more bandwidth than each of trading peers (were many more). Also I've got slightly different results with: Trading 75% vs Seeding 25% vs both enabled.

Gate in autolimiter also works OK for me, but again I always used reasonable total incoming/outgoing limits. It is expected to see Gate 20% higher than your maximum upload bandwidth (with total limit turned off ofc) because:

I don't know if this post makes any sense to you, sorry for my English.

if not choked peers are indeed counted towards seeding bandwidth, then this could be part of my problem,
because there are more unchoked peers on download with 200 peers, than there are peers on my private uploads

counting them as trading would certainly help, but

-------------------------------------------------------------------------------------------------------------

i did quick test

upload limit 400, trading limit 10%, same priority

1 download with 1 trading peer (download and upload speed)
10 seeding transfers with 1-2 connected uploading peers

=> downloading has between 20-200 upload
=> seeding takes rest

if i manually limit upload on downloading transfer to 10, seeding takes rest and i stay on 400

why is % limit not respected? it should stay on 0-40 if its working

so not choked counting towards seed is not only problem of this limit, it simply does not do, it was designed for?

Try to inverse your Outgoing Seeding/Trading % setting. If you have Trading 10% use Seeding 90% instead (with trading unselected). Then try both selected. See what happens.

does nothing, tested every possible setup, even if i go from 5 to 95 both or only one option, nothing in bandwidth graph changes... is simply does not work

not gonna bother with this anymore, they should fix it

It works OK for me on every version since 1.72 on both Windows and Linux. Perhaps some other setting causes this problem or it's a bug. I could test your settings if you are willing to send me the configuration. I could also send you mine if you'd like. If it's a bug, it would help developers to fix it if we knew when the Bandwidth Seeding/Trading % doesn't work.

gave it last change, downloaded portable version, run it with default settings

used 2 torrents
one is my current upload with no other seeds, this will represent many seeding peers
second is public torrent with 5k swarm, to get enough trading peers

result ... same as before even with setting it to 50/50 (to avoid possibly bad description) downloading transfer used most of the upload bandwidth anyway

=> its not settings, so back to manually limiting every new public download, not much else I can do to make it work

I tried to make Tixati to behave as you described. I succeeded with global Limit Outgoing KB/s turned off or set above my maximum upload bandwidth. If I set hard upload limit to some lower value (the graph is flat) it works OK.

Instead limiting manually every public download try using Categories as a workaround. It would be a bit easier for you to limit whole "public" category to some upload value and add public transfers to that category. You can also set Capture options for that category, so Tixati would automatically add public transfers to "public" category.