Re: Use of Filter rules for news, slow

On Thu, 26 Apr 2007, Lucio Chiappetti wrote:
> On Tue, 24 Apr 2007, Ross wrote:
>> On Mon, 23 Apr 2007, Lucio Chiappetti wrote:
> [about a filter rule on "News: news.group.name" being much slower than a
> rule on "From: address" when filtering from a folder list of newsgroups]

>> Did you try different "Current Folder Type" selections?
>> According to Pine Help: "When reading a newsgroup, there may be a performance
>> penalty incurred when collecting the information necessary to check whether or
>> not a Pattern matches a message. For this reason, the default Current Folder
>> Type is set to "Email". ...snip

> 1) I am not talking of screen redrawing, but of the initial folder
>   (newsgroup) opening which is slow

I refer not to the second half of the help paragraph, just the 
first concerning newsgroup performance in collecting information 
by pattern matching. I originally copied & pasted the entire 
paragraph. I now snip away the irrelevant, see above. However;

> 2) My "Current Folder Type" is not set to "News" but to a specific
>   Folder List = {newservername/nntp}#news.it.groupname
>   It will make no sense to apply my rule to all e-mail or to
>   all news, it has to be applied only to a couple of newsgroups !

I responded incorrectly to your question then, you desire to use 
the "Folder List" design.

> 3) the first rule I apply to the same newsgroup ("delete" messages
>   from two particular users, i.e. implementing a kill file) is fast.
>   It is only the rule which tests the Newsgroup header to contain
>   a specific newsgroup (i.e. to "delete" messages addressed to
>   group A (in Folder List) and crossposted to group B (unwanted))
>   which is slow.

Does the first rule not test the newsgroups header? If the first 
rule is pattern matching "From:" like you suggest "from two 
particular users", then pattern matching of the newsgroup is the 
same for each filter, just different patterns of the header. Both 
filters should be of similar speed, save something I am not aware 
of.

If pattern matching of the first message is not being processed, 
it would be faster. Only the delete flag which is really outside 
the message would be checked, not a pattern match process.

> 4) I thought that defining a pattern of "News" (when the header field
>   is called Newsgroups) might have some problems when fetching the
>   messages. Maybe pine is trying to fetch all messages to compare,
>   while it should fetch only those marked unread or not deleted in
>   .newsrc for the specific newsgroup (A).
>   I tried changing it to an "extra header pattern" Newsgroups, but
>   still it is slow.

I tend to believe, based on postings here of experienced users & 
my use, the "unread" flag is almost useless for this type 
application, where the "deleted" flag check or "not deleted" 
filter selection works very well. However;

Many messages posted here, including from you I remember, 
historically indicate newsgroups are slower to process & filter 
as Pine help indicates, & Pine is not the best solution. Though I 
must accept procmail cannot help with this one or you would use 
it.

Not knowing how many filters you operate, if substantial, 
possibly filter sequence would help. Once many messages were 
deleted, there would be much less to process for any particular 
filter. I filter junkmail almost next to the last. The good stuff 
is picked out first. But I do not process newsgroups.

--
The Best to You & Yours,
RossARR
[email protected]
mrossarr.nixsyspaus.org/
---end-of-message---
_______________________________________________
Pine-info mailing list
[email protected]
http://mailman1.u.washington.edu/mailman/listinfo/pine-info