#41 Rate implementation doubled

2.9.0 branch
closed-fixed
nobody
Remailer (20)
5
2003-05-03
2003-01-29
Anonymous
No

In mix.cfg I have:

RATE 10

In the pool I have, for example, 67 messages. Then I do:

./mix -dS

This adds one dummy message to the pool, for a total of 68
and flushes the pool leaving 56 messages behind.

If I repeat ./mix -dS a few more times I see the following
number of messages left behind in the pool:

47, 40, 34, 29, 25

Doing the math:

(68-56)/56 = .21
(57-47)/47 = .21
(48-40)/40 = .20
(35-29)/29 = .21
(30-25)/25 = .20

Conclusion:

RATE is implemented to be twice what is set in mix.cfg file.
Applies to client usage as well.

Discussion

  • Logged In: YES
    user_id=1525

    I cannot reproduce this.
    weasel@valiant:~/Mix-cvs$ for i in `seq 1 20`; do /bin/ls -1
    pool/m* | wc -l; ./mix -dS >/dev/null 2>&1; done
    100
    91
    83
    76
    70
    64
    59
    54
    50
    46
    43
    40
    37
    35
    33
    31
    29
    27
    26
    25

     
    • status: open --> open-works-for-me
     
  • Logged In: NO

    Well, here is the result on the same experiment here on Red Hat
    Linux 7.3. In mix.cfg I have RATE 10 Once it gets to 18 it
    behaves properly.

    [mixer@monoscope Mix]$ for i in `seq 1 20`; do /bin/ls -1
    pool/m* | wc -l; ./mix -dS >/dev/null 2>&1; done
    100
    82
    68
    57
    48
    41
    35
    30
    26
    23
    20
    18
    17
    16
    15
    14
    13
    12
    11
    10

     
  • Logged In: YES
    user_id=1525

    are there any other options in your mix.cfg file? If yes,
    can you please paste the entire file in here? And which
    version to you run exactly? 2.9.0 or an earlier beta?

    TIA

     
  • Logged In: NO

    Mixmaster Version 2.9.0

    ----------Contents of mix.cfg----------
    SMTPRELAY localhost
    REMAIL n
    HELONAME lamer.com
    CHAIN *,*,*,*
    NUMCOPIES 1
    DISTANCE 2
    MINREL 98
    RELFINAL 99
    MAXLAT 2h
    RATE 10
    MAILtoNEWS mail2news@anon.lcs.mit.edu
    ERRLOG /var/log/mixlog

     
  • Logged In: NO

    (Hope this doesn't get posted twice)

    Mixmaster 2.9.0

    mix.cfg file:

    SMTPRELAY localhost
    REMAIL n
    HELONAME lamer.com
    CHAIN *,*,*,*
    NUMCOPIES 1
    DISTANCE 2
    MINREL 98
    RELFINAL 99
    MAXLAT 2h
    RATE 10
    MAILtoNEWS mail2news@anon.lcs.mit.edu
    ERRLOG /var/log/mixlog

     
  • Logged In: NO

    Some additional info on this.

    Use the ncurses interface to generate 100 dummy messages by
    hammering the 'd' key.

    Then begin pressing the 's' key.

    The number of messages in the pool decreases properly (~10%
    with RATE 10) with each press of the 's' key.

     
  • Logged In: NO

    I've determined that 2.9.0, rc1, and beta-41 also have this
    problem. Beta-40 is OK, so maybe that will offer a clue.

     
  • Logged In: YES
    user_id=1525

    please set SENDPOOLTIME to 0h:

    SENDPOOLTIME 0h

    is the problem still reproduceable with beta40 now?

    [one of the changes between them was to set the default to 0]
    [anyway, I got an idea - will post later]

     
  • Logged In: YES
    user_id=1525

    I meant "does the problem still not show in b40"

     
  • Logged In: YES
    user_id=1525

    weasel@valiant:~/Mix-cvs$ for i in `seq 1 20`; do /bin/ls -1
    pool/m* | wc -l; ./mix -dS >/dev/null 2>&1; done
    100
    82
    68
    57
    48
    41
    35
    30
    26
    23
    20
    18
    17
    16

    Ha!

    If remail is NO, we always do a pool run.
    if you run mix -M and sendpooltime is low enogh (i.e. 0)
    it's done twice.

    [removed the works-for-me]

     
  • Logged In: NO

    Yes, with SENDPOOLTIME 0h in beta 40, the bad behavior
    shows up.

    I'll remove SENDPOOLTIME 0h and just sit with beta 40 for a
    while while you look into it further.

    Thanks for your time and patience.

     
  • Logged In: YES
    user_id=1525

    setting SENDPOOLTIME to 1h or similar will make 2.9.0 behave
    similar enough to what you want.

    I need to think and talk to Len and Ulf about how to fix this.

     
  • Logged In: NO

    OK, thanks for the workaround. I'll give it a try. Beta 40 works
    fine for me too, so I can always just stick there.

     
    • status: open-works-for-me --> closed-works-for-me
     
    • status: closed-works-for-me --> closed-fixed
     
  • Logged In: YES
    user_id=1525

    Fixed in CVS. new option: clientautoflush