There are a few things I don't like about this implementation, such as starting the feeCutoff at 0 (instead of 1000 sat/kb). But I do think an approach like this would work to "solve" this problem in 0.10 and 0.11 backports. The problem with #6673 is there was no limit to how high the fee went, but if its capped at some value that is still potentially reasonable for a min relay fee (like in this pull), then it might be better than just changing a hard coded value.

Starting the feeCutoff at 0 is easily changed, (we can start it anywhere
) but it does rise quite quickly once the mempool is full. When you
compile and run use debug=mempool and the log entries show you how the
values change as the mempool increases or decreases.

On 10/10/2015 2:50 PM, Alex Morcos wrote:

There are a few things I don't like about this implementation, such as
starting the feeCutoff at 0 (instead of 1000 sat/kb). But I do think
an approach like this would work to "solve" this problem in 0.10 and
0.11 backports. The problem with #6673
#6673 is there was no limit
to how high the fee went, but if its capped at some value that is
still potentially reasonable for a min relay fee (like in this pull),
then it might be better than just changing a hard coded value.

—
Reply to this email directly or view it on GitHub
#6803 (comment).

I really prefer #6722 over this. It seems gratuitous to have a less-effecient implementation for mempool limiting unless we decide #6722 is overcomplicated or otherwise not mergedable.

Oh I meant in consideration for a backport not for master.

Sent from my iPhone

On Oct 14, 2015, at 6:04 AM, Matt Corallo [email protected] wrote:

I really prefer #6722 over this. It seems gratuitous to have a less-effecient implementation for mempool limiting unless we decide #6722 is overcomplicated or otherwise not mergedable.

—
Reply to this email directly or view it on GitHub.

Here's a plot of the last 12 hours showing the actual mempool size and cumulative blocked (spam) transactions over time (x axis in minutes, y axis in MiB).

Closing this because I think it's been eclipsed by the mempool limiting thats merged and the related fixes around it in the pipeline. If I've made an error, feel free to yell at me. :)