If you read the case that sai himself opened, you'll see that the
Spring config generates a subclass of LinkedHashMap which is enough to
preserve order of execution.

So, for theI users wiring the server via Spring there's no need to
specify a Map version.  For embedders invoking the API methos
directly, I agree with Niklas that we shouldn't force them to use
LinkedHashMap (what if they want to use a SortedMap?) or any other map
implementation, execution order here is the responsability of the
user, not ours.

This said, Sai has a valid point when he speaks about predictability
and I could be led to agree with him  if he still insisted on that
option. But, if javadoc is OK I don't think this change is really
needed, and would mean breaking our API so it would be available just
for the 'trunk' version.


2011/3/18 Christian Gosch <[email protected]>:
> Hmm...
>
> To refer to the closed issue, it would be helpful if the example Spring
> configuration named "config-full.xml" would contain a ftplets element
> which *does* define a Map type to use -- but I cannot see anything like
> this inside this file (see the attached files, as of v1.0.5,
> 2010-SEP-26, 6:24PM).
>
>
> --cg
>
>> -----Original Message-----
>> From: Sai Pullabhotla [mailto:[email protected]]
>> Sent: Friday, March 18, 2011 2:09 PM
>> To: [email protected]
>> Subject: Re: FTPlet entrySet: Sort order for multiple active FTPlets?
>>
>> Well... there was a open case about this, which is now closed...
>>
>> https://issues.apache.org/jira/browse/FTPSERVER-223
>>
>> Are you sure this is still an issue?
>>
>> On Fri, Mar 18, 2011 at 7:46 AM, Christian Gosch
>> <[email protected]> wrote:
>> > Hi,
>> >
>> > I just looked at the implementation of
>> > DefaultFtpletContainer.onConnect() and saw that it processes all
>> > contained (registered) Ftplets by traversing the (concurrent hash)
> map
>> > of declared Ftplets, just as onDisconnect().
>> >
>> > But it does so based on the ftplets.entrySet() and the sort order
> which
>> > it imposes on the entry set, which in turn is "undefined" in that it
>> > does not guarantee any special sort order.
>> >
>> > Why is the concurrent map ftplets not implemented as a map with a
>> > reliable sort order depending on the key values?
>> >
>> > It may be good practice to have every registered Ftplet act
>> > independently of any other in the same container, but there may be
> good
>> > reasons to have an ordered sequence, may be by order of declaration,
> or
>> > by order of key or whatever may be appropriate.
>> >
>> > Is there a special rationale behind this?
>> >
>> >
>> > btw: I did not yet find any time to play around with this really :-(
>> >
>> >
>> > Regards,
>> > --
>> > Dipl.-Inform. Christian Gosch, PMI PMP
>> > Systems Architecture, Project Management
>> >
>> > inovex GmbH
>> > Büro Pforzheim
>> > Karlsruher Strasse 71
>> > D-75179 Pforzheim
>> > Tel: +49 (0)7231 3191-85
>> > Fax: +49 (0)7231 3191-91
>> > [email protected]
>> > www.inovex.de
>> >
>> > Sitz der Gesellschaft: Pforzheim
>> > AG Mannheim, HRB 502126
>> > Geschäftsführer: Stephan Müller
>> >
>> >
>> >
>> >
>>
>> !DSPAM:4d83598117594243614118!
>>
>>
>
>
>

Reply via email to