so how can I test a modified gate_news?
I've got this gate_news script modified to run SpamBayes but I'm unsure how I can test it. I'm fairly confident that I got most things right, but the only news-to-mail gateway I have access to is for comp.lang.python, not one I'd care to hose up too badly.
It occurs to me that maybe I could set up a one-way gateway from some arbitrary low-traffic newsgroup to a one-off mailing list. I haven't any experience setting up gate_news. Can someone here suggest a recipe for testing this beast which doesn't require too much scaffolding?
Thanks,
-- Skip Montanaro - skip@pobox.com - http://smontanaro.dyndns.org/
on 11/20/08 11:06 PM, skip@pobox.com said:
It occurs to me that maybe I could set up a one-way gateway from some arbitrary low-traffic newsgroup to a one-off mailing list. I haven't any experience setting up gate_news. Can someone here suggest a recipe for testing this beast which doesn't require too much scaffolding?
The issue here is that you not only need to set up a different version of gate_news, you really kind of need to set up a different copy of Mailman to play with, and you need a few test e-mail addresses to use as your mail-side subscribers, and a news server account somewhere else in the world so that you can see what news articles look like when they go out.
Any of the standard "test" newsgroups would be fine to use as your guinea pig, since you're going to go ahead and throw all that stuff away when you're done.
Barry does a lot of this kind of testing in his own little playground on the machines he owns, and I believe that Mark does the same.
We can see about setting up a second copy of Mailman (and gate_news) on the python.org machines, but it would be a little easier to test with if we could do that on a machine that did not already have an existing production Mailman installation.
-- Brad Knowles <brad@shub-internet.org> LinkedIn Profile: <http://tinyurl.com/y8kpxu>
On 11/21/2008 12:18 AM, Brad Knowles wrote:
The issue here is that you not only need to set up a different version of gate_news, you really kind of need to set up a different copy of Mailman to play with, and you need a few test e-mail addresses to use as your mail-side subscribers, and a news server account somewhere else in the world so that you can see what news articles look like when they go out.
I had not thought about the separate news server. Though I don't completely understand why you can't test on the same news server.
We can see about setting up a second copy of Mailman (and gate_news) on the python.org machines, but it would be a little easier to test with if we could do that on a machine that did not already have an existing production Mailman installation.
(Refer to my last email.)
If you will let me know what you want installed for your test bed, I'll set something up. I can also provide email addresses for testing if you don't care about the domain that they are at. Just let me know.
Grant. . . .
on 11/21/08 12:24 AM, Grant Taylor said:
I had not thought about the separate news server. Though I don't completely understand why you can't test on the same news server.
You could actually test on the same news server for sending out the traffic, but you still need that independent verification test, to see how things look once they get off that local server.
You need to do more than just the loopback test, although the loopback test is where you start.
If you will let me know what you want installed for your test bed, I'll set something up. I can also provide email addresses for testing if you don't care about the domain that they are at. Just let me know.
We'll keep that in mind. And if Skip wants to move forward on setting this up on a schedule that is tough for the members of the postmaster team to assist with, then of course he's certainly free to contact you for that assistance.
That's a decision I'll leave up to him.
-- Brad Knowles <brad@shub-internet.org> LinkedIn Profile: <http://tinyurl.com/y8kpxu>
On 11/20/2008 11:06 PM, skip@pobox.com wrote:
I've got this gate_news script modified to run SpamBayes but I'm unsure how I can test it. I'm fairly confident that I got most things right, but the only news-to-mail gateway I have access to is for comp.lang.python, not one I'd care to hose up too badly.
Aw, why not. :P
It occurs to me that maybe I could set up a one-way gateway from some arbitrary low-traffic newsgroup to a one-off mailing list. I haven't any experience setting up gate_news. Can someone here suggest a recipe for testing this beast which doesn't require too much scaffolding?
Why not set up another mailing list that does the same gatewaying but this time through the SpamBayes filter.
As I sit here and type that I'm betting that the filtering will be Mailman install wide, not just mailing list specific. (Presuming that this is the case.) Why not shut down your Mailman instance and copy it to another location for testing. Go ahead and re-start your first instance and then tweak the second one for the testing in a new location. I think that would allow you to run tow completely independent Mailman installs with different features.
This is presuming that Mailman will allow its self to run multiple instances on the same system. If this is not possible, let me know as I have a system that I could do a temporary install of Mailman and do some testing and compare it against another install that I have.
Though, something to keep in mind is that now is not really the best time to try things with the drop in spam with McColo being shut down and all...
Thanks,
*nod*
Grant. . . .
on 11/21/08 12:18 AM, Grant Taylor said:
Why not set up another mailing list that does the same gatewaying but this time through the SpamBayes filter.
As I sit here and type that I'm betting that the filtering will be Mailman install wide, not just mailing list specific. (Presuming that this is the case.)
Correct. That would be site-wide.
Why not shut down your Mailman instance and copy itto another location for testing.
I'd kind of rather not shut it down unless there isn't any other option. We do hundreds of thousands of e-mail messages per day, and I'd rather not disrupt any of the other 100+ production lists on the server unless there is no other choice.
Go ahead and re-start your firstinstance and then tweak the second one for the testing in a new location. I think that would allow you to run tow completely independent Mailman installs with different features.
Mailman can do that, yes. But it takes some more work, and you can't just blindly copy one mailman installation to another and then expect to start the second one up without a lot of additional work.
If you want to run multiple copies of Mailman on the same server, you really kind of need to do fresh installs for each additional copy. This makes sure that all the paths are set correctly for each copy, including which different configuration file they are built to use by default, which different directory structures they're configured to use, etc....
This is presuming that Mailman will allow its self to run multiple instances on the same system.
It can be done, but it takes a bit more work.
If this is not possible, let me know as Ihave a system that I could do a temporary install of Mailman and do some testing and compare it against another install that I have.
If Skip wants to go that direction and doesn't have a machine of his own that he can use for this function, I think that the PSF has another couple of machines around that we could probably use for some short-term testing.
In that case, we'll just need to make sure that the appropriate accounts are in place, etc.... And one or more members of the postmaster team will probably need to be involved in helping set that stuff up, so there's going to be a bit of coordination to be done.
Though, something to keep in mind is that now is not really the best time to try things with the drop in spam with McColo being shut down and all...
In this case, we're looking at USENET spam being gatewayed to e-mail, and I have no evidence that there has been any appreciable dropoff in USENET spam over the last couple of weeks. So, this should still be a valid test going that direction.
-- Brad Knowles <brad@shub-internet.org> LinkedIn Profile: <http://tinyurl.com/y8kpxu>
participants (3)
-
Brad Knowles -
Grant Taylor -
skip＠pobox.com