rtable presenting, present. Those who prefer writing,
write.
--
Nick Kew
his separation, which is what
makes it
a great platform for applications involving content filtering. Stream-editor
seeks to
offer basic content filtering for ATS, but is limited by the lack of such clean
separation.
--
Nick Kew
> On 7 May 2018, at 12:49, Leif Hedstrom wrote:
>
> Hi all,
>
> as discussed today, we’d like to propose the following plugins to be promoted
> from experimental to stable:
On a similar note, how about the stream_editor?
Anyone want to confess to operational usage?
—
Nick Kew
uses
ironbee's core classes and APIs and avoids the need for std.
--
Nick Kew
e to use Ironbee for your purposes without
any new plugin.
--
Nick Kew
t to any origin
> server.
Ironbee does that, including both the basics and workarounds
for a number of glitches. Might be worth taking something
from it as a startingpoint.
--
Nick Kew
On Mon, 22 Feb 2016 12:44:05 -0700
Leif Hedstrom wrote:
> Scheduled: Feb 23, 2016, 10:00 AM to 10:30 AM
Timezone?
--
Nick Kew
that's been made and is being passed down.
Did I miss on-list discussion leading to it?
--
Nick Kew
hat symbol?
Does compiling with -O0 make any difference?
--
Nick Kew
(gdb)
I suspected this might be an artifact of the build process, but
I get the same result building stream_editor.so by hand.
Any ideas?
--
Nick Kew
ncidental?
This comes back to my bugbear of API review. If we have
containers and tools to do similar things, we should expose
them in an API, and with a stability promise.
The exercise could be a beneficial review, too.
--
Nick Kew
e-phase callback to determine
Continue/Error when the rendezvous's condition is satisfied.
Does that make any sense?
--
Nick Kew
" button. Plain text, please!
--
Nick Kew
e need to parse (or at least pattern-match)
an HTTP request line to detect HTTP.
If starting from scratch I'd do both. That is to say, allow
ports to be configured to a protocol (and drop/reject anything
inappropriate thereunto), but also detect the request line in
data before firing a TXN_START event.
--
Nick Kew
gh in the wrong order in my opinion, but Alan says that this is a
> known issue).
I've seen both orders. I agree it would be much easier if we could
rely on seeing TXN_CLOSE first, but it can be worked around.
--
Nick Kew
duce exactly that on wheezy. The culprit is the
--enable-static-proxy, without which it builds just fine.
What's the history of that option? I've never used it,
and it appears not to exist in 4.x.
--
Nick Kew
and establish yourself
firmly in the Web of Trust?
--
Nick Kew
scussion by projecting an IRC window
alongside the speaker's presentation, so we can say something
and get noticed. Alternative IRC-based variants might also help.
Also it would be good to be able to see those presentation slides
that aren't already referenced from the wiki!
--
Nick Kew
blank screen.
--
Nick Kew
in servers to desktop browsers - implemented this
cacheing scheme, you'd still lose MOST of the benefits of
cacheing, as the same content passes through different paths.
--
Nick Kew
to offer
any specific advice. You could make the AV a direct plugin to ATS,
though that would likely involve re-inventing some of what Ironbee
provides for you.
--
Nick Kew
ec and depending on that produce or cancel and
>> then send a redirect. What do you think?
I should certainly hope you could do it better and more efficiently with
ironbee.
At least, assuming there's an API for your proposed virus scanner!
--
Nick Kew
ose cousin to two
crashes we've encountered recently in filter functions. Maybe the fix
suggested in the thread at
http://comments.gmane.org/gmane.comp.apache.trafficserver.devel/1920
might be worth a try?
--
Nick Kew
On 15 Aug 2014, at 21:55, James Peach wrote:
> What's a real use case for iterating over all the headers?
Ironbee. Or any WAF, or WAF framework.
Or performance analysis tools.
--
Nick Kew
stroy you're
looking for. It's just the one MIME header!
--
Nick Kew
On Fri, 20 Jun 2014 21:55:19 +0100
Nick Kew wrote:
> Patch attached.
Bah. It seems the listserv stripped the patch.
Uploaded instead:
http://people.apache.org/~niq/patches/null-transform.patch
--
Nick Kew
ionally the same as the change that fixed ironbee.
This is actually a slightly bigger change, because I had to introduce
things that didn't exist before, as opposed to just move them around!
The change is to give the TSVConn the lifetime of the tx.
Patch attached.
--
Nick Kew
T.
2. Destroy it at the end of the txn
in the handler for TS_EVENT_HTTP_TXN_CLOSE.
This simple lifetime change fixes the crash! I wonder if it would
be Good Practice to make an equivalent change in the bundled
transforms?
--
Nick Kew
rced
into a choice which, as you say, nothing should be forcing.
Or else our plugin takes on diverging versions. One for HTTPD,
one for nginx, one for trafficserver-old, one for trafficserver-new?
--
Nick Kew
o support both before- and after- TS
versions. Can I make a plea to avoid that: maybe a new function name,
and migrate the old one as a #define wrapper for the new? That applies
equally to the above with flags or to just adding the c-l-len argument.
BTW, we use it with NULL content-type argument! I'm not about to
dig deep in github to remind myself of the evolution of it,
but I guess it's because we only use text/html and set it elsewhere.
--
Nick Kew
rivial utility to something interesting!
--
Nick Kew
f discussions on the topic of making it possible to
> reload plugins on a "live" system.
Isn't that a separate issue? A plugin can register an alternative-init
function that works as a reload-init with TSMgmtUpdateRegister,
just subject to TS-2036.
--
Nick Kew
e in
> plugins/gzip/gzip.cc, plugins/experimental/esi/esi.cc and
> lib/atscppapi/src/TransformationPlugin.cc.
That seems to answer a slightly different question. You're still (from the
Client's perspective) processing normally, but just returning an error
on completion.
--
Nick Kew
eveloper wouldn't
have to ask)!
Crossposting to dev@ where at least my question belongs.
--
Nick Kew
it self-aware, but would that have to
involve anything more than traffic_line -x on a timer?
--
Nick Kew
On Fri, 13 Sep 2013 22:43:22 +0800
永豪 wrote:
> ## RFC: how to do file monitor and config file reload
How would you expect this to relate to "traffic_line -x"?
--
Nick Kew
On 7 Sep 2013, at 01:35, Nick Kew wrote:
>> Something in these commits breaks the builds completely. I have both expat
>> and libxml2 installed (with dev packages), and it doesn't detect either.
>> Same problem (I think) on all build bots (they all fail as far as I can
libxml2) is required.
Damn. Guess we could use some temporary debug so you get
a trace of what it tried.
I should find time to revisit it over the weekend.
--
Nick Kew
take a look at what I did with APR, which was
a more complex job than TS!
--
Nick Kew
partial document
tree. But I'm not volunteering to do that anytime soon.
--
Nick Kew
diff --git a/mgmt/stats/StatProcessor.cc b/mgmt/stats/StatProcessor.cc
index 268dc29..22d1ff9 100644
--- a/mgmt/stats/StatProcessor.cc
+++ b/mgmt/stats/StatProcessor.cc
@@ -65,7 +65,8 @@ startElement(void * /* userDat
if it's only a laptop with
good quality telephony and a VOIP (or skype) address.
p.s. apologies if another mail just got sent from here. This laptop
is out of control and needs pensioning off.
--
Nick Kew
On 15 Jul 2013, at 23:49, Leif Hedstrom wrote:
> Hi all,
>
> so one of the outcomes from the latest BarCamp, is that there might be a
> place for one, or even two, official ATS Summits. The ideas floating was
> something like:
>
> Release Summit
> --
>
> * To be held in the June
logic as yours, only to find side-effects. Bottom line seems to be,
yes you can and indeed must move _writer and lose its previous
block in certain operations. You'd better have a reader already
allocated or you'll never again see that block!
--
Nick Kew
On 24 Apr 2013, at 15:43, Leif Hedstrom wrote:
> On 4/24/13 6:51 AM, Nick Kew wrote:
>> On Fri, 19 Apr 2013 22:54:47 +0100
>> Nick Kew wrote:
>>
>>
>>> I'll attach a mini demo-plugin that reproduces the problem.
>>> The essence of it can be b
On Fri, 19 Apr 2013 22:54:47 +0100
Nick Kew wrote:
> I'll attach a mini demo-plugin that reproduces the problem.
> The essence of it can be boiled down to:
Magic sauce here.
This fails:
TSHttpTxnClientReqGet(txnp, &bufp, &hdr_loc);
iobufp = TSIOBufferCreate();
hat
a single extra byte in the header makes the difference
between working fine and provoking the problem.
A bug or a PBKAC?
[Bcc: my colleague]
--
Nick Kew
#include
#include
#include
#include
static void process_hdr(TSHttpTxn txnp)
{
TSMBuffer bufp;
TSIOBuffer iobufp;
TSIOB
On Wed, 17 Apr 2013 09:16:00 -0600
Leif Hedstrom wrote:
> On 4/17/13 5:37 AM, Nick Kew wrote:
> > On Mon, 15 Apr 2013 16:35:35 -0600
> > Leif Hedstrom wrote:
> >
> >> The goal is to release this on April 18th, so please cast your +, - or 0
> >> vo
th).
Looks like it needs a LONG_MAX construct. Or an explicitly-sized type.
Surely I can't be the only one on 32 bits?
Manually fixing those constants, all else is well.
It passes make test, and runs as a drop-in update and is
binary-compatible with my existing ironbee binaries.
--
Nick Kew
e, its atexit would also run first, while the library was still
loaded.
--
Nick Kew
On Fri, 05 Apr 2013 09:10:47 -0700
James Peach wrote:
> BTW, why does it bother trying to unload the libraries?
A sense of tidiness and order?
There's precious little in my life, but it's Good Practice
in software, IMHO.
--
Nick Kew
http://mail-archives.apache.org/mod_mbox/trafficserver-dev/201112.mbox/%3C974f0e63-35f5-466d-8fc4-5b0482177baa@iris%3E
--
Nick Kew
lists!
> I have a tiny patchset which switches from the old default
> latin1 to the utf-8 up on my github:
>
> https://github.com/igalic/trafficserver/tree/utf-8
An actual patch would be easier to review.
Your "utf-8" changes appear to span more than one commit.
--
Nick Kew
l contribution would
need IP clearance from yourselves and your clients.
Details at www.apache.org.
Having said all that, if you are indeed making a significant
contribution in a tricky area, that's great and welcome.
Looking forward to it!
--
Nick Kew
builds fine, passes make check
on Linux and MacOSX.
I'll make that a firm +1 if it works well with my plugins
(or if any problems arising appear to be on my side).
But that can wait 'til tomorrow.
--
Nick Kew
In fact, I may do that today, if I'm spared my semi-regular
Sunday evening commitment.
--
Nick Kew
The APIs for working with HTTP headers appear to be designed
to support multi-valued (comma-separated) headers. So as I
understand it, I should be able to do things like:
nval = TSMimeHdrFieldValuesCount(...);
for (i=0; i
t; somehow inspect the video itself? What is the API?)
You would use a filter for that. The null-filter example provides a template
for passing data through, so you'd just want to add code to inspect as much
as you want of it.
--
Nick Kew
On Tue, 12 Jun 2012 16:51:42 -0600
Leif Hedstrom wrote:
> On 6/11/12 11:20 PM, Nick Kew wrote:
> > On 12 Jun 2012, at 02:49, Leif Hedstrom wrote:
> >
> >> Another option that I've been pondering, which is more suitable for large
> >> contents I think, is t
ould be overkill. I wonder if there's a case
for adding a heuristic to examine the client's ranges, and fetch
the whole thing while the client waits UNLESS the number of
bytes the client wants to skip exceeds some threshold - which
then triggers what you describe?
--
Nick Kew
a (as in either a
remap or a filter), you don't yet have them in a TXN_START function.
--
Nick Kew
ts,
> and annotate the commits (and jira tickets) accordingly.
Seems OK, except insofar as it begs the question of whether we want to
push everyone into creating jira issues even where they may seem pointless.
Sounds like you're basically saying jira should be a public record of
every non-trivial commit. I can live with that, but would just like to
confirm whether that's what you mean?
--
Nick Kew
r confusion, and
version numbers are cheap!
--
Nick Kew
plies exactly where it's needed,
or whether it also affects cases where a transform is configured
but operates before rather than after caching?
--
Nick Kew
t on cache hits the transformation runs on the
> cached untransformed object and we don't see the chunked encoding.
Sorry, can't help you with that one. Is it getting returned with a
content-length
that's invalidated by your transform? What happens if your transform unsets it?
--
Nick Kew
bee/blob/master/servers/trafficserver/ironbee.c
Not entirely. There's nothing async about the ironbee plugin!
It just processes events in order using the standard TS model.
--
Nick Kew
dependent opinion might tell us something worthwhile about
just how useful the API really is as of now.
--
Nick Kew
browse/TS-1008
--
Nick Kew
proposed patch appears to want to change away from that!
--
Nick Kew
On 9 Feb 2012, at 11:57, Nick Kew wrote:
> I'll post my patch here so as not to delay it while I find my way
> into ASF git.
Looks like the listserver stripped out the attachment. Inlining ...
Index: STATUS
===
#x27;m guessing that's down to the git
migration? I'll post my patch here so as not to delay it while I find my way
into ASF git.
More reviews/votes may or may not be forthcoming in/near time for your
release schedule.
--
Nick Kew
sday, I might have a good session
reviewing some of those. Time (if any) before then is more likely to be
limited.
--
Nick Kew
On 6 Feb 2012, at 14:45, Nick Kew wrote:
> Getting this on MacOS:
Followup: I compiled with -O0 in CFLAGS and CXXFLAGS, and with --enable-debug
The problem goes away and it runs OK.
--
Nick Kew
ve
tried on it. However, 3.0 works. Will see if turning off gcc optimisation
makes any difference.
--
Nick Kew
est and I abandoned it. I have the impression that ESI
is just used by a handful of orgs, each with its own implementation.
--
Nick Kew
On Tue, 20 Dec 2011 15:02:05 -0500
Jason Giedymin wrote:
> It sounds like your leaving!?
Who's leaving? He's posting in the persona of
> > -- Traffic Server Community
Whoops!
--
Nick Kew
e any publically-visible pre-apache code history?
--
Nick Kew
Ironbee, and that my
boss has indicated he'd be happy to donate.
Basic synopsis: an equivalent of httpd's LoadFile to
load libraries and export their symbols.
I'll attach it. Any interest?
--
Nick Kew
/
f classes (like map, bag, set).
No grief from gcc updates nor even other-cc since then.
But if you have a good use case where templates serve best, ...
--
Nick Kew
tc/fedora-release ]; then
DISTRIB_ID="fedora"
elif [ -e /etc/redhat-release ]; then
... [more of same]
So for an amazon-hosted node I was able to work around it with
a symlink from /etc/system-release to /etc/redhat-release.
Is there a reason for this test, or is it just a historic legacy?
--
Nick Kew
ble with. One point raised on httpd
is that early adopters should preferably be prepared to put some
effort in to improving the infrastructure. Working in python on
git hooks was specifically mentioned.
--
Nick Kew
s were observed with no plugin; the
zero Content-Length appears to be triggered by a plugin.
--
Nick Kew
Cache-Control: max-age=60^M
Age: 2^M
^M
WARNING: Response code not 2xx (304)
According to ab, this is in response to:
GET /testfile.html HTTP/1.0^M
Host: localhost^M
User-Agent: ApacheBench/2.3^M
Accept: */*^M
^M
i.e. an unconditional request!
This looks infeasibly broken. Any clue what I'm missing?
--
Nick Kew
a role really adds anything. I can't recollect a [PATCH] being
posted to this list, and if it does happen I think our natural reaction
(certainly mine) would already be to suggest the poster opens
a ticket in the issue tracker.
But if you'd like to take explicit responsibility that's fine!
--
Nick Kew
ially different dev team
or release cycle to the base product. But that would have
to be decided on a case-by-case basis, not a one-size-fits-all,
let alone an indiscriminate bundling!
--
Nick Kew
e not necessarily the same as in reverse_map.
--
Nick Kew
Just tried to commit a fix to TS-1008 (low-hanging fruit)
in trunk. I'm getting 403 (Forbidden) from svn.
I was prepared for a possible 401, but whence 403?
Anyway, patch attached in case anyone else would care to
commit and fares better!
--
Nick Kew
Index: proxy/InkA
On 26 Oct 2011, at 10:58, Igor Galić wrote:
>
> We already have a TS API component, of course. :)
Heh. Is that new, or did I just miss it? I recollect seeing Management API
but not TS API in the list. Bah.
Thanks anyway.
--
Nick Kew
ested improvements?
--
Nick Kew
ust for shit and giggles, what behaviour do you get if you use this
> hook to read the request header?
The Host: header is indeed as sent by the Client.
The Request line isn't.
GET /path/
has become
GET http:///path/
(yep, three consecutive slashes)
--
Nick Kew
o actual forward proxy is involved!
Apart from these apparently-inconsistent semantics being confusing,
the version I'm looking for is the first one, and I don't seem able to get it!
Is this a bug?
--
Nick Kew
On Tue, 04 Oct 2011 09:49:46 -0600
Leif Hedstrom wrote:
> On 10/04/2011 07:11 AM, Nick Kew wrote:
> > I'd like to write a plugin that works on TCP Connection data.
> > In HTTPD I'd just work with the conn_rec and write
> > connection-level filters.
> >
>
es while the latter will see
all data. TS's HTTP_[REQUEST|RESPONSE]_TRANSFORM work
only at request level and don't see HTTP headers,
request lines, status lines, or stray data (e.g.
malformed requests/responses). Or what am I missing?
--
Nick Kew
#x27;s something
> we can sensibly solve via TSPluginMechanisms()
FWIW, I just (since posting) hacked up a new plugin whose
sole purpose is to implement that.
Now to see whether it solves my original problem!
--
Nick Kew
alisation (like traffiserver's
TSPluginInit).
I'd like to add an equivalent function and directive
to trafficserver. Any comments and/or objections?
--
Nick Kew
se of a bogus pointer in the library. Doesn't this cause
a lot of brittleness in production?
--
Nick Kew
a time
convenient to everyone interested, so where's the problem?
If that were to prove impossible, that's a bridge to cross when we
come to it, whether through a compromise like variable times,
or through abandoning the idea altogether. But let's not dismiss
a medium just because of what *might* happen!
--
Nick Kew
thers in Europe (hi Igor!) are fine with much later
hours - our time. On a Thursday I could do six hours later for
instance, if attendance after a couple of pints is acceptable!
Anyway, don't let my absence become a showstopper if finding
an alternative slot proves a challenge.
--
Nick Kew
On 16 Nov 2010, at 00:37, Leif Hedstrom wrote:
>Where: #ts-weekly-dev on irc.freenode.net (standard IRC server for Apache)
>When: Every Thursday at 10am PST
That time of day/week precludes me, as I'm en route to a weekly event at 6pm
UK time on a Thursday.
--
Nick Kew
On 8 Nov 2010, at 01:29, Leif Hedstrom wrote:
> On 11/07/2010 06:21 PM, Nick Kew wrote:
>> A while back I put httpd's mod_proxy through the coadvisor test suite[1] and
>> fixed a
>> bunch of protocol violations, most of them pedantic edge-cases.
>>
>>
trafficserver.
[1] http://www.measurement-factory.com/
--
Nick Kew
1 - 100 of 114 matches
Mail list logo
