You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@tomcat.apache.org by Martin Algesten <pu...@taglab.com> on 2002/10/30 12:19:14 UTC
production quality?
Hi all,
Just some thoughts.
I've been using the 3.3.1 release for quite some time in a
mod_jk/apache/linux kind of setup and all was fine. Though a couple of
weeks ago I felt a need to start looking at new versions of all my
API's/products in order to make sure I stay on top of things and don't
end up with unsupported versions.
What do we mean with production quality?
According to the Tomcat project home page, 4.1.12 is a production
quality release, however using it in real life makes me question the
usefulness of such status. I've been monitoring this list and also tried
to contribute by discussing/submitting patches for the bugs I've
encountered. I don't have an issue with how long it takes to resolve
these issues, after all we are all doing this for fun (more or less ;)
). However I do think we have a responsibility in what signals we're
sending regarding how useful a release really is. The current 4.1.12
release have some quite nasty issues that in some production setups
makes it more or less useless. In my opinion the most nasty issues are
those that directly breaks internet standards and the core API (10373,
13846, 13040).
What about quality control?
Well, the bugzilla do allow for a Severity, Priority classification.
Perhaps we should start classifying bugs more actively using these
switches. This could be combined with a policy that we don't make a
release before certain levels of issues are all resolved? I also get a
feeling that we as developers are somewhat in the wrong position to
label things as "production quality". The watchdog tests are all very
well, but can never replace real life scenarios. Perhaps we need a new
label such as release candidate? Perhaps we should be more in the spirit
of other open source projects, (OpenSSL and Mozilla comes to mind),
where the release cycle seems to more involve end users and have a more
cautious labelling of releases.
This is not intended as criticism of anyone. After all Tomcat is a
fantastic project with fantastic people contributing. Good work all!
Martin
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: production quality?
Posted by Martin Algesten <ma...@taglab.com>.
The reports I referred to were just the ones affecting me in my setup,
and my point was not about their validity, rather just an overall
feeling from using this particular software.
I have not tested other containers for spec compatibility, I can't claim
to have even tested this one.
I have a number of smaller websites with various amounts of traffic
running on Tomcat 3.3.1, and yes the uptime is remarkably good. Software
is never perfect and any setup will work differently. I don't expect to
be able to change my 3.3.1 to 4.1.12 and have everything working without
any effort. Different live scenarios always implies different
workarounds/tradeoffs. So in that sense you're either using a particular
subset of functionality that works in your live scenario or you've
worked around any issues that you encountered.
Regards,
Martin
Glenn Nielsen wrote:
> Martin Algesten wrote:
>
>> Hi all,
>>
>> Just some thoughts.
>>
>> I've been using the 3.3.1 release for quite some time in a
>> mod_jk/apache/linux kind of setup and all was fine. Though a couple of
>> weeks ago I felt a need to start looking at new versions of all my
>> API's/products in order to make sure I stay on top of things and don't
>> end up with unsupported versions.
>>
>> What do we mean with production quality?
>> According to the Tomcat project home page, 4.1.12 is a production
>> quality release, however using it in real life makes me question the
>> usefulness of such status. I've been monitoring this list and also tried
>> to contribute by discussing/submitting patches for the bugs I've
>> encountered. I don't have an issue with how long it takes to resolve
>> these issues, after all we are all doing this for fun (more or less ;)
>> ). However I do think we have a responsibility in what signals we're
>> sending regarding how useful a release really is. The current 4.1.12
>> release have some quite nasty issues that in some production setups
>> makes it more or less useless. In my opinion the most nasty issues are
>> those that directly breaks internet standards and the core API (10373,
>> 13846, 13040).
>>
>
> I quickly reviewed the three bug reports above. All three seem to be
> related to spec compliance and interpretation of the spec.
>
> Just curious, have you done the same level of spec compliance testing
> with other conatiners? How well do they comply?
>
> For me use Tomcat 4.1 is production quality. I use it with Apache
> mod_jk/AJP
> on a medium volume site with 20-30K tomcat requests per day with peaks
> of over
> 5000 tomcat requests per hour. And I get uptimes of over four weeks.
>
> Regards,
>
> Glenn
>
>
> --
> To unsubscribe, e-mail:
> <ma...@jakarta.apache.org>
> For additional commands, e-mail:
> <ma...@jakarta.apache.org>
>
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: production quality?
Posted by Glenn Nielsen <gl...@mail.more.net>.
Martin Algesten wrote:
> Hi all,
>
> Just some thoughts.
>
> I've been using the 3.3.1 release for quite some time in a
> mod_jk/apache/linux kind of setup and all was fine. Though a couple of
> weeks ago I felt a need to start looking at new versions of all my
> API's/products in order to make sure I stay on top of things and don't
> end up with unsupported versions.
>
> What do we mean with production quality?
> According to the Tomcat project home page, 4.1.12 is a production
> quality release, however using it in real life makes me question the
> usefulness of such status. I've been monitoring this list and also tried
> to contribute by discussing/submitting patches for the bugs I've
> encountered. I don't have an issue with how long it takes to resolve
> these issues, after all we are all doing this for fun (more or less ;)
> ). However I do think we have a responsibility in what signals we're
> sending regarding how useful a release really is. The current 4.1.12
> release have some quite nasty issues that in some production setups
> makes it more or less useless. In my opinion the most nasty issues are
> those that directly breaks internet standards and the core API (10373,
> 13846, 13040).
>
I quickly reviewed the three bug reports above. All three seem to be
related to spec compliance and interpretation of the spec.
Just curious, have you done the same level of spec compliance testing
with other conatiners? How well do they comply?
For me use Tomcat 4.1 is production quality. I use it with Apache mod_jk/AJP
on a medium volume site with 20-30K tomcat requests per day with peaks of over
5000 tomcat requests per hour. And I get uptimes of over four weeks.
Regards,
Glenn
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: production quality?
Posted by Martin Algesten <pu...@taglab.com>.
Yup.. as I said, generally content-type changes might be ok. And yes I
have reported it as a bug.
Martin
jean-frederic clere wrote:
> Martin Algesten wrote:
>
>> I don't see how chaning the Content-Type can be right in this
>> scenario (yes, generally content-type changes might be ok), can you
>> explain. If I understood this correctly we are asking Tomcat if a
>> specific object has changed (If-Modified-Since), and any response
>> should surely be true about the resource we're asking about. E.g. if
>> we're asking if a certain gif file has changed since yesterday, then
>> tomcat tells us that it is unchanged content wise, but meta-wise it
>> is no longer a gif, but an html document?!
>
>
> Yes, in the scenario that is a bug. So report it as a bug.
>
> But think of an ISO image incorrectly transferted as text/html due to
> _bad_ mime type configuration, would not it be nice to just change the
> mime type and not resend the whole file?
>
>>
>> Martin
>>
>>
>>
>> jean-frederic clere wrote:
>>
>>> Martin Algesten wrote:
>>>
>>>> 1. Tomcat should either not send any headers on a 304, or if it
>>>> does then make sure that they do reflect the correct values for the
>>>> requested object (e.g. not call a gif a text).
>>>
>>>
>>>
>>>
>>> Changing a Content-Type is legitimate, but changing the
>>> Content-Length is probably not.
>>>
>>> But I do not think that Tomcat has to change the Content-Type.
>>>
>>>>
>>>> 2. Not entirely sure here. Reading the HTTP/1.1
>>>> http://www.w3.org/Protocols/rfc2616/rfc2616-sec10.html#sec10.3.5
>>>> It states:
>>>> "If a cache uses a received 304 response to update a cache entry,
>>>> the cache MUST update the entry to reflect any new field values
>>>> given in the response."
>>>>
>>>> I suspect that is what we're seeing in mod_proxy. But I just
>>>> realised that I have some reading to do, the stuff about weak and
>>>> strong validator didn't make much sense to me...
>>>
>>>
>>>
>>>
>>> It changes the Content-Type and I think that is OK.
>>>
>>>>
>>>> Martin
>>>>
>>>> jean-frederic clere wrote:
>>>>
>>>>> Pier Fumagalli wrote:
>>>>>
>>>>>> On 30/10/02 20:02, "Martin Algesten" <ma...@taglab.com> wrote:
>>>>>>
>>>>>>
>>>>>>> In a nutshell mod_proxy updates its cached entries with whatever new
>>>>>>> headers are given to it. E.g. first request comes into mod_proxy
>>>>>>> and it
>>>>>>> can't find the requested resource in its cache. It forwards on to my
>>>>>>> tomcat who responds with something like:
>>>>>>> HTTP/1.1 200
>>>>>>> Content-Type: image/gif
>>>>>>> Content-Length: 12345
>>>>>>>
>>>>>>> Second call comes into mod_proxy this time with an
>>>>>>> "If-Modified-Since"
>>>>>>> for the same resource. mod_proxy needs to revalidate its cached
>>>>>>> entry
>>>>>>> against tomcat and does an "If-Modifed-Since" against tomcat and
>>>>>>> tomcat
>>>>>>> answers:
>>>>>>> HTTP/1.1 304
>>>>>>> Content-Type: text/html
>>>>>>> Content-Length: 0
>>>>>>>
>>>>>>> At this point mod_proxy updates it's cached entry and ends up
>>>>>>> with a gif
>>>>>>> that has got a Content-Type set to text/html.
>>>>>>>
>>>>>>> Further requests to mod_proxy without "If-Modified-Since" results in
>>>>>>> GIFs with strange content types. Thank god for IE not trusting the
>>>>>>> content type :)
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> Nope, that's not it, but it's a good catch. We don't keep proxied
>>>>>> content
>>>>>> cached... Thanks a lot for the clarification...
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> If I got it right there are 2 errors:
>>>>> 1 - Tomcat should not send a Content-Type nor Content-Length.
>>>>> 2 - mod_proxy should complain because we are sending "garbages" or
>>>>> ignore the "invalid" headers.
>>>>>
>>>>>>
>>>>>> Pier (gone diggin' mirrors)
>>>>>>
>>>>>>
>>>>>> --
>>>>>> To unsubscribe, e-mail:
>>>>>> <ma...@jakarta.apache.org>
>>>>>> For additional commands, e-mail:
>>>>>> <ma...@jakarta.apache.org>
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> --
>>>>> To unsubscribe, e-mail:
>>>>> <ma...@jakarta.apache.org>
>>>>> For additional commands, e-mail:
>>>>> <ma...@jakarta.apache.org>
>>>>>
>>>>
>>>>
>>>> --
>>>> To unsubscribe, e-mail:
>>>> <ma...@jakarta.apache.org>
>>>> For additional commands, e-mail:
>>>> <ma...@jakarta.apache.org>
>>>>
>>>>
>>>
>>>
>>>
>>>
>>>
>>> --
>>> To unsubscribe, e-mail:
>>> <ma...@jakarta.apache.org>
>>> For additional commands, e-mail:
>>> <ma...@jakarta.apache.org>
>>>
>>
>>
>> --
>> To unsubscribe, e-mail:
>> <ma...@jakarta.apache.org>
>> For additional commands, e-mail:
>> <ma...@jakarta.apache.org>
>>
>>
>
>
>
>
> --
> To unsubscribe, e-mail:
> <ma...@jakarta.apache.org>
> For additional commands, e-mail:
> <ma...@jakarta.apache.org>
>
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: production quality?
Posted by jean-frederic clere <jf...@fujitsu-siemens.com>.
Martin Algesten wrote:
> I don't see how chaning the Content-Type can be right in this scenario
> (yes, generally content-type changes might be ok), can you explain. If I
> understood this correctly we are asking Tomcat if a specific object has
> changed (If-Modified-Since), and any response should surely be true
> about the resource we're asking about. E.g. if we're asking if a certain
> gif file has changed since yesterday, then tomcat tells us that it is
> unchanged content wise, but meta-wise it is no longer a gif, but an html
> document?!
Yes, in the scenario that is a bug. So report it as a bug.
But think of an ISO image incorrectly transferted as text/html due to _bad_ mime
type configuration, would not it be nice to just change the mime type and not
resend the whole file?
>
> Martin
>
>
>
> jean-frederic clere wrote:
>
>> Martin Algesten wrote:
>>
>>> 1. Tomcat should either not send any headers on a 304, or if it does
>>> then make sure that they do reflect the correct values for the
>>> requested object (e.g. not call a gif a text).
>>
>>
>>
>> Changing a Content-Type is legitimate, but changing the
>> Content-Length is probably not.
>>
>> But I do not think that Tomcat has to change the Content-Type.
>>
>>>
>>> 2. Not entirely sure here. Reading the HTTP/1.1
>>> http://www.w3.org/Protocols/rfc2616/rfc2616-sec10.html#sec10.3.5
>>> It states:
>>> "If a cache uses a received 304 response to update a cache entry, the
>>> cache MUST update the entry to reflect any new field values given in
>>> the response."
>>>
>>> I suspect that is what we're seeing in mod_proxy. But I just realised
>>> that I have some reading to do, the stuff about weak and strong
>>> validator didn't make much sense to me...
>>
>>
>>
>> It changes the Content-Type and I think that is OK.
>>
>>>
>>> Martin
>>>
>>> jean-frederic clere wrote:
>>>
>>>> Pier Fumagalli wrote:
>>>>
>>>>> On 30/10/02 20:02, "Martin Algesten" <ma...@taglab.com> wrote:
>>>>>
>>>>>
>>>>>> In a nutshell mod_proxy updates its cached entries with whatever new
>>>>>> headers are given to it. E.g. first request comes into mod_proxy
>>>>>> and it
>>>>>> can't find the requested resource in its cache. It forwards on to my
>>>>>> tomcat who responds with something like:
>>>>>> HTTP/1.1 200
>>>>>> Content-Type: image/gif
>>>>>> Content-Length: 12345
>>>>>>
>>>>>> Second call comes into mod_proxy this time with an
>>>>>> "If-Modified-Since"
>>>>>> for the same resource. mod_proxy needs to revalidate its cached entry
>>>>>> against tomcat and does an "If-Modifed-Since" against tomcat and
>>>>>> tomcat
>>>>>> answers:
>>>>>> HTTP/1.1 304
>>>>>> Content-Type: text/html
>>>>>> Content-Length: 0
>>>>>>
>>>>>> At this point mod_proxy updates it's cached entry and ends up with
>>>>>> a gif
>>>>>> that has got a Content-Type set to text/html.
>>>>>>
>>>>>> Further requests to mod_proxy without "If-Modified-Since" results in
>>>>>> GIFs with strange content types. Thank god for IE not trusting the
>>>>>> content type :)
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> Nope, that's not it, but it's a good catch. We don't keep proxied
>>>>> content
>>>>> cached... Thanks a lot for the clarification...
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> If I got it right there are 2 errors:
>>>> 1 - Tomcat should not send a Content-Type nor Content-Length.
>>>> 2 - mod_proxy should complain because we are sending "garbages" or
>>>> ignore the "invalid" headers.
>>>>
>>>>>
>>>>> Pier (gone diggin' mirrors)
>>>>>
>>>>>
>>>>> --
>>>>> To unsubscribe, e-mail:
>>>>> <ma...@jakarta.apache.org>
>>>>> For additional commands, e-mail:
>>>>> <ma...@jakarta.apache.org>
>>>>>
>>>>>
>>>>
>>>>
>>>>
>>>>
>>>> --
>>>> To unsubscribe, e-mail:
>>>> <ma...@jakarta.apache.org>
>>>> For additional commands, e-mail:
>>>> <ma...@jakarta.apache.org>
>>>>
>>>
>>>
>>> --
>>> To unsubscribe, e-mail:
>>> <ma...@jakarta.apache.org>
>>> For additional commands, e-mail:
>>> <ma...@jakarta.apache.org>
>>>
>>>
>>
>>
>>
>>
>>
>> --
>> To unsubscribe, e-mail:
>> <ma...@jakarta.apache.org>
>> For additional commands, e-mail:
>> <ma...@jakarta.apache.org>
>>
>
>
> --
> To unsubscribe, e-mail:
> <ma...@jakarta.apache.org>
> For additional commands, e-mail:
> <ma...@jakarta.apache.org>
>
>
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: production quality?
Posted by Martin Algesten <ma...@taglab.com>.
I don't see how chaning the Content-Type can be right in this scenario
(yes, generally content-type changes might be ok), can you explain. If I
understood this correctly we are asking Tomcat if a specific object has
changed (If-Modified-Since), and any response should surely be true
about the resource we're asking about. E.g. if we're asking if a certain
gif file has changed since yesterday, then tomcat tells us that it is
unchanged content wise, but meta-wise it is no longer a gif, but an html
document?!
Martin
jean-frederic clere wrote:
> Martin Algesten wrote:
>
>> 1. Tomcat should either not send any headers on a 304, or if it does
>> then make sure that they do reflect the correct values for the
>> requested object (e.g. not call a gif a text).
>
>
> Changing a Content-Type is legitimate, but changing the
> Content-Length is probably not.
>
> But I do not think that Tomcat has to change the Content-Type.
>
>>
>> 2. Not entirely sure here. Reading the HTTP/1.1
>> http://www.w3.org/Protocols/rfc2616/rfc2616-sec10.html#sec10.3.5
>> It states:
>> "If a cache uses a received 304 response to update a cache entry, the
>> cache MUST update the entry to reflect any new field values given in
>> the response."
>>
>> I suspect that is what we're seeing in mod_proxy. But I just realised
>> that I have some reading to do, the stuff about weak and strong
>> validator didn't make much sense to me...
>
>
> It changes the Content-Type and I think that is OK.
>
>>
>> Martin
>>
>> jean-frederic clere wrote:
>>
>>> Pier Fumagalli wrote:
>>>
>>>> On 30/10/02 20:02, "Martin Algesten" <ma...@taglab.com> wrote:
>>>>
>>>>
>>>>> In a nutshell mod_proxy updates its cached entries with whatever new
>>>>> headers are given to it. E.g. first request comes into mod_proxy
>>>>> and it
>>>>> can't find the requested resource in its cache. It forwards on to my
>>>>> tomcat who responds with something like:
>>>>> HTTP/1.1 200
>>>>> Content-Type: image/gif
>>>>> Content-Length: 12345
>>>>>
>>>>> Second call comes into mod_proxy this time with an "If-Modified-Since"
>>>>> for the same resource. mod_proxy needs to revalidate its cached entry
>>>>> against tomcat and does an "If-Modifed-Since" against tomcat and
>>>>> tomcat
>>>>> answers:
>>>>> HTTP/1.1 304
>>>>> Content-Type: text/html
>>>>> Content-Length: 0
>>>>>
>>>>> At this point mod_proxy updates it's cached entry and ends up with
>>>>> a gif
>>>>> that has got a Content-Type set to text/html.
>>>>>
>>>>> Further requests to mod_proxy without "If-Modified-Since" results in
>>>>> GIFs with strange content types. Thank god for IE not trusting the
>>>>> content type :)
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> Nope, that's not it, but it's a good catch. We don't keep proxied
>>>> content
>>>> cached... Thanks a lot for the clarification...
>>>
>>>
>>>
>>>
>>> If I got it right there are 2 errors:
>>> 1 - Tomcat should not send a Content-Type nor Content-Length.
>>> 2 - mod_proxy should complain because we are sending "garbages" or
>>> ignore the "invalid" headers.
>>>
>>>>
>>>> Pier (gone diggin' mirrors)
>>>>
>>>>
>>>> --
>>>> To unsubscribe, e-mail:
>>>> <ma...@jakarta.apache.org>
>>>> For additional commands, e-mail:
>>>> <ma...@jakarta.apache.org>
>>>>
>>>>
>>>
>>>
>>>
>>>
>>> --
>>> To unsubscribe, e-mail:
>>> <ma...@jakarta.apache.org>
>>> For additional commands, e-mail:
>>> <ma...@jakarta.apache.org>
>>>
>>
>>
>> --
>> To unsubscribe, e-mail:
>> <ma...@jakarta.apache.org>
>> For additional commands, e-mail:
>> <ma...@jakarta.apache.org>
>>
>>
>
>
>
>
>
> --
> To unsubscribe, e-mail:
> <ma...@jakarta.apache.org>
> For additional commands, e-mail:
> <ma...@jakarta.apache.org>
>
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: production quality?
Posted by jean-frederic clere <jf...@fujitsu-siemens.com>.
Martin Algesten wrote:
> 1. Tomcat should either not send any headers on a 304, or if it does
> then make sure that they do reflect the correct values for the requested
> object (e.g. not call a gif a text).
Changing a Content-Type is legitimate, but changing the
Content-Length is probably not.
But I do not think that Tomcat has to change the Content-Type.
>
> 2. Not entirely sure here. Reading the HTTP/1.1
> http://www.w3.org/Protocols/rfc2616/rfc2616-sec10.html#sec10.3.5
> It states:
> "If a cache uses a received 304 response to update a cache entry, the
> cache MUST update the entry to reflect any new field values given in the
> response."
>
> I suspect that is what we're seeing in mod_proxy. But I just realised
> that I have some reading to do, the stuff about weak and strong
> validator didn't make much sense to me...
It changes the Content-Type and I think that is OK.
>
> Martin
>
> jean-frederic clere wrote:
>
>> Pier Fumagalli wrote:
>>
>>> On 30/10/02 20:02, "Martin Algesten" <ma...@taglab.com> wrote:
>>>
>>>
>>>> In a nutshell mod_proxy updates its cached entries with whatever new
>>>> headers are given to it. E.g. first request comes into mod_proxy and it
>>>> can't find the requested resource in its cache. It forwards on to my
>>>> tomcat who responds with something like:
>>>> HTTP/1.1 200
>>>> Content-Type: image/gif
>>>> Content-Length: 12345
>>>>
>>>> Second call comes into mod_proxy this time with an "If-Modified-Since"
>>>> for the same resource. mod_proxy needs to revalidate its cached entry
>>>> against tomcat and does an "If-Modifed-Since" against tomcat and tomcat
>>>> answers:
>>>> HTTP/1.1 304
>>>> Content-Type: text/html
>>>> Content-Length: 0
>>>>
>>>> At this point mod_proxy updates it's cached entry and ends up with a
>>>> gif
>>>> that has got a Content-Type set to text/html.
>>>>
>>>> Further requests to mod_proxy without "If-Modified-Since" results in
>>>> GIFs with strange content types. Thank god for IE not trusting the
>>>> content type :)
>>>
>>>
>>>
>>>
>>> Nope, that's not it, but it's a good catch. We don't keep proxied
>>> content
>>> cached... Thanks a lot for the clarification...
>>
>>
>>
>> If I got it right there are 2 errors:
>> 1 - Tomcat should not send a Content-Type nor Content-Length.
>> 2 - mod_proxy should complain because we are sending "garbages" or
>> ignore the "invalid" headers.
>>
>>>
>>> Pier (gone diggin' mirrors)
>>>
>>>
>>> --
>>> To unsubscribe, e-mail:
>>> <ma...@jakarta.apache.org>
>>> For additional commands, e-mail:
>>> <ma...@jakarta.apache.org>
>>>
>>>
>>
>>
>>
>>
>> --
>> To unsubscribe, e-mail:
>> <ma...@jakarta.apache.org>
>> For additional commands, e-mail:
>> <ma...@jakarta.apache.org>
>>
>
>
> --
> To unsubscribe, e-mail:
> <ma...@jakarta.apache.org>
> For additional commands, e-mail:
> <ma...@jakarta.apache.org>
>
>
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: production quality?
Posted by Pier Fumagalli <pi...@betaversion.org>.
On 31/10/02 14:36 "Martin Algesten" <ma...@taglab.com> wrote:
> 1. Tomcat should either not send any headers on a 304, or if it does
> then make sure that they do reflect the correct values for the requested
> object (e.g. not call a gif a text).
>
> 2. Not entirely sure here. Reading the HTTP/1.1
> http://www.w3.org/Protocols/rfc2616/rfc2616-sec10.html#sec10.3.5
> It states:
> "If a cache uses a received 304 response to update a cache entry, the
> cache MUST update the entry to reflect any new field values given in the
> response."
>
> I suspect that is what we're seeing in mod_proxy. But I just realised
> that I have some reading to do, the stuff about weak and strong
> validator didn't make much sense to me...
I too believe that mod_proxy is behaving quite correctly according to the
spec... Tomcat should not send those headers... It will also impact client
(browser) caches, not only mod_proxy...
Pier
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: production quality?
Posted by Martin Algesten <ma...@taglab.com>.
1. Tomcat should either not send any headers on a 304, or if it does
then make sure that they do reflect the correct values for the requested
object (e.g. not call a gif a text).
2. Not entirely sure here. Reading the HTTP/1.1
http://www.w3.org/Protocols/rfc2616/rfc2616-sec10.html#sec10.3.5
It states:
"If a cache uses a received 304 response to update a cache entry, the
cache MUST update the entry to reflect any new field values given in the
response."
I suspect that is what we're seeing in mod_proxy. But I just realised
that I have some reading to do, the stuff about weak and strong
validator didn't make much sense to me...
Martin
jean-frederic clere wrote:
> Pier Fumagalli wrote:
>
>> On 30/10/02 20:02, "Martin Algesten" <ma...@taglab.com> wrote:
>>
>>
>>> In a nutshell mod_proxy updates its cached entries with whatever new
>>> headers are given to it. E.g. first request comes into mod_proxy and it
>>> can't find the requested resource in its cache. It forwards on to my
>>> tomcat who responds with something like:
>>> HTTP/1.1 200
>>> Content-Type: image/gif
>>> Content-Length: 12345
>>>
>>> Second call comes into mod_proxy this time with an "If-Modified-Since"
>>> for the same resource. mod_proxy needs to revalidate its cached entry
>>> against tomcat and does an "If-Modifed-Since" against tomcat and tomcat
>>> answers:
>>> HTTP/1.1 304
>>> Content-Type: text/html
>>> Content-Length: 0
>>>
>>> At this point mod_proxy updates it's cached entry and ends up with a gif
>>> that has got a Content-Type set to text/html.
>>>
>>> Further requests to mod_proxy without "If-Modified-Since" results in
>>> GIFs with strange content types. Thank god for IE not trusting the
>>> content type :)
>>
>>
>>
>> Nope, that's not it, but it's a good catch. We don't keep proxied
>> content
>> cached... Thanks a lot for the clarification...
>
>
> If I got it right there are 2 errors:
> 1 - Tomcat should not send a Content-Type nor Content-Length.
> 2 - mod_proxy should complain because we are sending "garbages" or
> ignore the "invalid" headers.
>
>>
>> Pier (gone diggin' mirrors)
>>
>>
>> --
>> To unsubscribe, e-mail:
>> <ma...@jakarta.apache.org>
>> For additional commands, e-mail:
>> <ma...@jakarta.apache.org>
>>
>>
>
>
>
>
> --
> To unsubscribe, e-mail:
> <ma...@jakarta.apache.org>
> For additional commands, e-mail:
> <ma...@jakarta.apache.org>
>
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: production quality?
Posted by jean-frederic clere <jf...@fujitsu-siemens.com>.
Pier Fumagalli wrote:
> On 30/10/02 20:02, "Martin Algesten" <ma...@taglab.com> wrote:
>
>
>>In a nutshell mod_proxy updates its cached entries with whatever new
>>headers are given to it. E.g. first request comes into mod_proxy and it
>>can't find the requested resource in its cache. It forwards on to my
>>tomcat who responds with something like:
>>HTTP/1.1 200
>>Content-Type: image/gif
>>Content-Length: 12345
>>
>>Second call comes into mod_proxy this time with an "If-Modified-Since"
>>for the same resource. mod_proxy needs to revalidate its cached entry
>>against tomcat and does an "If-Modifed-Since" against tomcat and tomcat
>>answers:
>>HTTP/1.1 304
>>Content-Type: text/html
>>Content-Length: 0
>>
>>At this point mod_proxy updates it's cached entry and ends up with a gif
>>that has got a Content-Type set to text/html.
>>
>>Further requests to mod_proxy without "If-Modified-Since" results in
>>GIFs with strange content types. Thank god for IE not trusting the
>>content type :)
>
>
> Nope, that's not it, but it's a good catch. We don't keep proxied content
> cached... Thanks a lot for the clarification...
If I got it right there are 2 errors:
1 - Tomcat should not send a Content-Type nor Content-Length.
2 - mod_proxy should complain because we are sending "garbages" or ignore the
"invalid" headers.
>
> Pier (gone diggin' mirrors)
>
>
> --
> To unsubscribe, e-mail: <ma...@jakarta.apache.org>
> For additional commands, e-mail: <ma...@jakarta.apache.org>
>
>
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: production quality?
Posted by Pier Fumagalli <pi...@betaversion.org>.
On 30/10/02 20:02, "Martin Algesten" <ma...@taglab.com> wrote:
> In a nutshell mod_proxy updates its cached entries with whatever new
> headers are given to it. E.g. first request comes into mod_proxy and it
> can't find the requested resource in its cache. It forwards on to my
> tomcat who responds with something like:
> HTTP/1.1 200
> Content-Type: image/gif
> Content-Length: 12345
>
> Second call comes into mod_proxy this time with an "If-Modified-Since"
> for the same resource. mod_proxy needs to revalidate its cached entry
> against tomcat and does an "If-Modifed-Since" against tomcat and tomcat
> answers:
> HTTP/1.1 304
> Content-Type: text/html
> Content-Length: 0
>
> At this point mod_proxy updates it's cached entry and ends up with a gif
> that has got a Content-Type set to text/html.
>
> Further requests to mod_proxy without "If-Modified-Since" results in
> GIFs with strange content types. Thank god for IE not trusting the
> content type :)
Nope, that's not it, but it's a good catch. We don't keep proxied content
cached... Thanks a lot for the clarification...
Pier (gone diggin' mirrors)
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: production quality?
Posted by Martin Algesten <ma...@taglab.com>.
In a nutshell mod_proxy updates its cached entries with whatever new
headers are given to it. E.g. first request comes into mod_proxy and it
can't find the requested resource in its cache. It forwards on to my
tomcat who responds with something like:
HTTP/1.1 200
Content-Type: image/gif
Content-Length: 12345
Second call comes into mod_proxy this time with an "If-Modified-Since"
for the same resource. mod_proxy needs to revalidate its cached entry
against tomcat and does an "If-Modifed-Since" against tomcat and tomcat
answers:
HTTP/1.1 304
Content-Type: text/html
Content-Length: 0
At this point mod_proxy updates it's cached entry and ends up with a gif
that has got a Content-Type set to text/html.
Further requests to mod_proxy without "If-Modified-Since" results in
GIFs with strange content types. Thank god for IE not trusting the
content type :)
Bug 13846 has my curl detective work in it...
M
Pier Fumagalli wrote:
>On 30/10/02 6:06 pm, "Martin Algesten" <ma...@taglab.com> wrote:
>
>
>
>>An incorrect If-Modified-Since response seems to shag mod_proxy. Now I
>>don't know how widely used mod_proxy is, or if perhaps this should be
>>fixed there instead. However I am concerned about this and as I pointed
>>out I am happy to look into it myself, however I would appreciate a hint
>>of what to look at. You mention a header refactoring, elaborate?
>>
>>
>
>Could it be what we are observing on http://nagoya.apache.org/eyebrowse when
>we pass stuff in the backend on port 8080?
>
>What would be the "outcome" of this? (I mean you know any way in which I can
>find out if this is the bug we're seeing)
>
> Pier
>
>
>--
>To unsubscribe, e-mail: <ma...@jakarta.apache.org>
>For additional commands, e-mail: <ma...@jakarta.apache.org>
>
>
>
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: production quality?
Posted by Pier Fumagalli <pi...@betaversion.org>.
On 30/10/02 6:06 pm, "Martin Algesten" <ma...@taglab.com> wrote:
> An incorrect If-Modified-Since response seems to shag mod_proxy. Now I
> don't know how widely used mod_proxy is, or if perhaps this should be
> fixed there instead. However I am concerned about this and as I pointed
> out I am happy to look into it myself, however I would appreciate a hint
> of what to look at. You mention a header refactoring, elaborate?
Could it be what we are observing on http://nagoya.apache.org/eyebrowse when
we pass stuff in the backend on port 8080?
What would be the "outcome" of this? (I mean you know any way in which I can
find out if this is the bug we're seeing)
Pier
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: production quality?
Posted by Martin Algesten <ma...@taglab.com>.
An incorrect If-Modified-Since response seems to shag mod_proxy. Now I
don't know how widely used mod_proxy is, or if perhaps this should be
fixed there instead. However I am concerned about this and as I pointed
out I am happy to look into it myself, however I would appreciate a hint
of what to look at. You mention a header refactoring, elaborate?
Martin
>> 13846: This is a minor bug (IMHO) which will need a rather risky
>> refactoring of the headers handling to fix, hence I decided to
>> postpone fixing it.
>>
>
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: production quality?
Posted by Martin Algesten <ma...@taglab.com>.
I'm dropping my argument about labelling now. It will probably not lead
anywhere anyway.
Martin
Remy Maucherat wrote:
> Martin Algesten wrote:
>
>> Hi all,
>>
>> Just some thoughts.
>>
>> I've been using the 3.3.1 release for quite some time in a
>> mod_jk/apache/linux kind of setup and all was fine. Though a couple of
>> weeks ago I felt a need to start looking at new versions of all my
>> API's/products in order to make sure I stay on top of things and don't
>> end up with unsupported versions.
>>
>> What do we mean with production quality?
>> According to the Tomcat project home page, 4.1.12 is a production
>> quality release, however using it in real life makes me question the
>> usefulness of such status. I've been monitoring this list and also tried
>> to contribute by discussing/submitting patches for the bugs I've
>> encountered. I don't have an issue with how long it takes to resolve
>> these issues, after all we are all doing this for fun (more or less ;)
>> ). However I do think we have a responsibility in what signals we're
>> sending regarding how useful a release really is. The current 4.1.12
>> release have some quite nasty issues that in some production setups
>> makes it more or less useless. In my opinion the most nasty issues are
>> those that directly breaks internet standards and the core API (10373,
>> 13846, 13040).
>
>
> 10373: I'm not convinced by the validity of the report.
> 13040: The spec is IMO not implementable, and most of the patches
> submitted were wrong. Hence, the issue is not fixed. Sorry.
> 13846: This is a minor bug (IMHO) which will need a rather risky
> refactoring of the headers handling to fix, hence I decided to
> postpone fixing it.
>
> Unfortunaltely, there's simply not enough developers to handle the
> amount of bug reports being filed. If the report is not crystal clear
> or is not easy to reproduce and investigate, then no one will look
> into it.
> The only way we could improve on this is that more people like you
> stepped in and helped fix bugs and investigate issues. Unfortunately,
> fixing bugs is not fun, and it doesn't happen very often (people
> prefer to complain about their favorite BZ report instead) :-(
>
> Remy
>
>
> --
> To unsubscribe, e-mail:
> <ma...@jakarta.apache.org>
> For additional commands, e-mail:
> <ma...@jakarta.apache.org>
>
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: production quality?
Posted by Martin Algesten <pu...@taglab.com>.
Hmm... perhaps not drop it after all. I had a think about it and I still
would argue that the amount of developers or the speed of turning out
fixes is not the issue. As I said before, we are all doing this more or
less for fun.
Taking on board that its all about expectations and that you need to
test your software before deploying it and that 4.1.12 can be
definitelly considered production quality for certain uses. There is
still a big difference between 3.3.1 and 4.1.12, the former seems to be
a more solid piece of software than the latter, and still they both
enjoy the same status on the Tomcat home page. Isn't just the sheer
amount of bug reports an argument in my direction?
Martin
Remy Maucherat wrote:
> Martin Algesten wrote:
>
>> Hi all,
>>
>> Just some thoughts.
>>
>> I've been using the 3.3.1 release for quite some time in a
>> mod_jk/apache/linux kind of setup and all was fine. Though a couple of
>> weeks ago I felt a need to start looking at new versions of all my
>> API's/products in order to make sure I stay on top of things and don't
>> end up with unsupported versions.
>>
>> What do we mean with production quality?
>> According to the Tomcat project home page, 4.1.12 is a production
>> quality release, however using it in real life makes me question the
>> usefulness of such status. I've been monitoring this list and also tried
>> to contribute by discussing/submitting patches for the bugs I've
>> encountered. I don't have an issue with how long it takes to resolve
>> these issues, after all we are all doing this for fun (more or less ;)
>> ). However I do think we have a responsibility in what signals we're
>> sending regarding how useful a release really is. The current 4.1.12
>> release have some quite nasty issues that in some production setups
>> makes it more or less useless. In my opinion the most nasty issues are
>> those that directly breaks internet standards and the core API (10373,
>> 13846, 13040).
>
>
> 10373: I'm not convinced by the validity of the report.
> 13040: The spec is IMO not implementable, and most of the patches
> submitted were wrong. Hence, the issue is not fixed. Sorry.
> 13846: This is a minor bug (IMHO) which will need a rather risky
> refactoring of the headers handling to fix, hence I decided to
> postpone fixing it.
>
> Unfortunaltely, there's simply not enough developers to handle the
> amount of bug reports being filed. If the report is not crystal clear
> or is not easy to reproduce and investigate, then no one will look
> into it.
> The only way we could improve on this is that more people like you
> stepped in and helped fix bugs and investigate issues. Unfortunately,
> fixing bugs is not fun, and it doesn't happen very often (people
> prefer to complain about their favorite BZ report instead) :-(
>
> Remy
>
>
> --
> To unsubscribe, e-mail:
> <ma...@jakarta.apache.org>
> For additional commands, e-mail:
> <ma...@jakarta.apache.org>
>
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: production quality?
Posted by Remy Maucherat <re...@apache.org>.
Martin Algesten wrote:
> Hi all,
>
> Just some thoughts.
>
> I've been using the 3.3.1 release for quite some time in a
> mod_jk/apache/linux kind of setup and all was fine. Though a couple of
> weeks ago I felt a need to start looking at new versions of all my
> API's/products in order to make sure I stay on top of things and don't
> end up with unsupported versions.
>
> What do we mean with production quality?
> According to the Tomcat project home page, 4.1.12 is a production
> quality release, however using it in real life makes me question the
> usefulness of such status. I've been monitoring this list and also tried
> to contribute by discussing/submitting patches for the bugs I've
> encountered. I don't have an issue with how long it takes to resolve
> these issues, after all we are all doing this for fun (more or less ;)
> ). However I do think we have a responsibility in what signals we're
> sending regarding how useful a release really is. The current 4.1.12
> release have some quite nasty issues that in some production setups
> makes it more or less useless. In my opinion the most nasty issues are
> those that directly breaks internet standards and the core API (10373,
> 13846, 13040).
10373: I'm not convinced by the validity of the report.
13040: The spec is IMO not implementable, and most of the patches
submitted were wrong. Hence, the issue is not fixed. Sorry.
13846: This is a minor bug (IMHO) which will need a rather risky
refactoring of the headers handling to fix, hence I decided to postpone
fixing it.
Unfortunaltely, there's simply not enough developers to handle the
amount of bug reports being filed. If the report is not crystal clear or
is not easy to reproduce and investigate, then no one will look into it.
The only way we could improve on this is that more people like you
stepped in and helped fix bugs and investigate issues. Unfortunately,
fixing bugs is not fun, and it doesn't happen very often (people prefer
to complain about their favorite BZ report instead) :-(
Remy
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>