Random GET requests instead of POST

[bcc1234]

Hey guys, I have a few forms that specify method="post", but about 5% of people who submit those forms send "get" requests without any parameters. Most seem to come from different versions of IE (including 5.x and 6.0).

I've searched usenet, but didn't find anything more than "yeah, that happens, not really sure why" posts.

The server side returns a 302 redirect once the form is submitted, but the problem occurs even before that.

Has anybody dealt with this before?

[jatar_k]

i have to admit i haven't seen that before, any other symptoms that you know of?

Have you managed to reproduce it?

[mcavic]

send "get" requests without any parameters.

I'd bet that this is a misbehaved spider, or more likely a spambot. Rather than bothering to figure out the form, it just grabs the action url hoping for some jucy content.

You might try putting a link to a nonexistant page, inside a comment block. If it's a spambot, I bet it'll try to grab that link too.

[bcc1234]

I'd bet that this is a misbehaved spider, or more likely a spambot.

And 10 minutes later that spambot sends me an e-mail stating that he was unable to place an order and asking for a phone number :)

[bcc1234]

jatar_k,
I know, it's really weird.

I could not find any similarities between occurrences.
All different ISP's, different browser versions (all IE, and a few visitors had Macs), different forms.

They are all forms in the same chain.
Like: enter billing addr, shipping addr, payment info...
And those thigs happen at different pages of the checkout chain.

I could not reproduce it. I also asked a lot of people to try placing test orders and everything worked fine.

[bcc1234]

Ok, a weird thing happened.
One visitor clicked a button (it's an image, not a submit button), his browser sent a GET request (and the error was displayed), then he hit the back button and clicked again - the second time it went through without any problems.
UA: Mozilla/4.0 (compatible; MSIE 5.0; Windows 98; DigExt)

Anybody? Any ideas? This is killing me!

[jatar_k]

I wish I had some answer, or hypothesis, for you but to be honest I've got nothing.

very strange

[rainborick]

I'm still seeing an unending stream of spammers attempting to locate open copies of Matt's FormMail using all sorts of variations on the file name and path. It wouldn't surprise me if some of the GETs you noticed were bots looking for the boilerplate HTML page that plain vanilla FormMail.pl spits out if you call it without any parameters by trying to execute the code from every <form> it finds.

[bcc1234]

rainborick, those are not bots. I'm 100% positive.

The last one was like this:
shopping cart page -> post
order detail form -> post
billing address form -> post
shipping address form -> get
error, back button
shipping address form -> post
payment info form -> post
submit order form -> post
order ok page

It could not have possibly been a bot, that placed an order and provided a cc :)

I'm not new to this whole thing, but this is the first time I've encountered something like this.

I even tried changing html for those forms
like
form action=".." method="POST"
or
form method="post" action=".."

no difference.

Also tried hitting enter instead of clicking the image with a mouse and many other things on different browsers. It always works ok.

[MonkeeSage]

This might a silly question, but are you showing the actual "GET ... HTTP/1.1" requests in your logs, or is it that your script is reading REQUEST_METHOD as GET, or something else?

Mabye some browser is garbling the "Content-length" on submission and thus interpreting the form as a GET because it can't send the correct Content-length header for a POST (like an internal fallback type thing)?

When it does the GET, do they see the full URL in the address bar with the?a=a&b=b&... syntax like they would if the form had a method="GET" attribute?

Just throwing out some thoughts in lue of an explanation. Very odd situation indeed!

Jordan

[bcc1234]

The script would accept both post and get without any problems and would treat them the same.
However, those requests don't have any parameters, just the file name. So it seems that browsers just drop the form alltogether and do a regular get as if this image was a link and not a button in the form.

[DrDoc]

Maybe if you make it a submit button instead?
I remember reading something somewhere a long time ago...
I'll try finding it.

[bcc1234]

DrDoc, I thought about it, but that won't look too good.

Besides, how many sites that use images in a form are there?
If that was the case, I'm sure it would be a known bug.

[DrDoc]

How many sites are using 'document.all' for their JavaScripts? But that doesn't mean it will work :)

Netscape 4 had well documented bugs regarding image submit buttons. And, if I'm not barking up the wrong tree, I believe there's something in IE5 as well...

[bcc1234]

I believe there's something in IE5 as well...

I tried searching for any references to that, but could not find anything about GET requests without parameters when POST should be made. Nothing solid, at least.

[grandpa]

bcc1234 - this is just a shot in the dark, but since no one else has provided a definite answer yet...

Is there something else happening at the page at the same time? I had a glitch on one of my pages the other day, not the same kind as we're talking about here, but it resulted from my uploading some changes while a user was busy shopping.

I wouldn't think that multiple users on the same page would ever cause this kind of problem, but what about a misbehaving spider?

Like I said, a shot in the dark. Something smells an awful lot like outside interference as opposed to a problem with the page. Just my opinion.

grandpa

[bcc1234]

grandpa, I thought about it, but it seems to happen on different pages. HTML seems to validate (except for the alt attributes).