Problems rendering an image + jpeg images not working

-- Tim Johnson <[tim--johnsons-web--com]> http://www.alaska-internet-solutions.com http://www.johnsons-web.com

Hi, Tim, I stole it from Gabriele!. ;-) I frankly don't remember where I first saw this kind of stuff, but it might have been an example in the rebol library. The option, by the way, is: t: read/binary %//rebol/view/nyc.jpg ... dat: copy "" foreach u t [append dat to-char u] set-modes system/ports/output [binary: true] write-io system/ports/output dat length? dat ... Who knows, maybe this version will work better. You can find out some more about get-modes and set-modes at: http://www.rebol.com/docs/core25.html I read several documents about the experiemental multipart mime, and played with it for about an hour over lunch. I could not get even a simple multipart text/html example to run properly on my Apache/REBOL/IE6 set-up on Win98. I am unsure why. I am out of time. Best of luck figuring this one out! --Scott Jones

Hi Scott:

Gabriele is a very good resource!

Me-thinks buffering is a better way to go, yes!

Thanks, that's going to be very useful.

That's a tough one all right (at least for my feeble brain). With Allen Kamp's help, I set up a feature to send html as an email message. I had a headache by the time I was finished. Thanks again... -- Tim Johnson <[tim--johnsons-web--com]> http://www.alaska-internet-solutions.com http://www.johnsons-web.com

I don't understand why you are doing it this why. Are you sure that using just: set-modes system/ports/output [binary: true] insert system/ports/output read/binary %image-file.jpg doesn't work? Regards, Gabriele. -- Gabriele Santilli <[g--santilli--tiscalinet--it]> -- REBOL Programmer Amigan -- AGI L'Aquila -- REB: http://web.tiscali.it/rebol/index.r

It works. My fault. Mea culpa. Ten thousand apologies. I could explain why I went off in that direction, but there is no learning benefit to it. So sorry. --Scott Jones

Scott I am still a bit confused by this thread. Please can you post the full working script and form ? thanks ./Jason

Me too! ;-)

Certainly. Of course, you will want to change your paths and pic names. #!/path/to/rebol --cgi -s REBOL [] print "Content-type: image/jpeg^/" set-modes system/ports/output [binary: true] insert system/ports/output read/binary %//rebol/view/nyc.jpg I creatively named the script "jpeg.cgi" and tossed it into my local website stuff in the cgi-bin directory (if you are on a *nix, then you will need to do a CHMOD to 755, or whatever), cranked up Apache, and cranked up IE6.0, then typed in the url: http://gsj/cgi-bin/jpeg.cgi What I personally could *not* get to work (or at least to work reliably) was the multipart/x-mixed-replace that Tim was seeking. I am beginning to wonder if, as Graham pointed out, that this may be more of a Internet Explorer problem than a REBOL or protocol problem. I don't currently have any other browser installed (he says with considerable embarrassment and wonder at why in the heck Microsoft became so insidiously and incredibly huge). It is important to note that the "x" in the mime name means experimental. I guess is it too experimental for Microsoft's more recent browsers to fully and reliably support. If there are any further questions, I would be happy to shed further confusion. ;-) --Scott Jones

My turn... :^) I believe the objective is to "stream" a series of images to the browser from a server (original source a web cam). The multipart/x-mixed-replace is a way to deliver to the browser multiple versions of the "same" image so that the browser should replace the image when it receives the next part in the multipart stream. IE apparently does not support this, for IE a activex object in the HTML page must be used. Theoretically "other" browsers might support it directly. My experience with Netscape so far shows promis because it does correctly finish with the final image, but does not show the earlier images. Netscape appears to be waiting until the connection is closed. The problem I think is that my REBOL script does not send any data to the browser until it quits. I've tried setting the mode of the output port to no-wait, but no change. Can someone solve this? Anyway, Tim, you'll probably have the ActiveX so this script might help you out by having something to compare with. I'm using three small gifs, but the principal should be the same as for the web cam. Things to watch: (1) Line endings. Note how I've set binary mode on for all my output. Therefore I need to supply the network standard line ending of CRLF. (2) Line endings again. Number of line endings between headers and content. Here's my work so far: --8<----- The HTML page --------------------- <html> <body> <h1>multipart/x-mixed-replace</h1> <center> <img border="1" src="http://localhost/cgi-bin/pushtest.r.cgi" height="37" width="37" alt="Streamed Image"> </center> </body> </html> --8<----- The CGI script --------------------- #!E:/Downloads/Languages/Rebol/Experimental/core2503031 -cs REBOL [ Title: {Server Push CGI Test} Author: {Brett Handley} Date: 29-Aug-2002 Comment: { See http://wp.netscape.com/assist/net_sites/pushpull.html } ] if error? err: try [ boundary-string: "ThisRandomString" boundary: rejoin [CRLF {--} boundary-string] emit: func [data] [ if block? data [data: rejoin data] insert system/ports/output data insert system/ports/output CRLF ] set-modes system/ports/output [binary: true] emit [ {Content-Type: multipart/x-mixed-replace;} { boundary-string="} boundary-string {"} ] emit boundary repeat i 3 [ file-data: read/binary file: join %n [i %.gif] emit {Content-Type: text/gif} emit {} emit file-data emit boundary write/append %pushtest.log join form now [ " " file " " length? file-data newline ] wait 00:00:01 ] ] [write/append %pushtest.log join form now [mold disarm err newline]] QUIT --8<----- Little script to test CGI --------------------- p: open/direct/binary/no-wait http://localhost/cgi-bin/pushtest.r.cgi until [ wait p none? if not none? s: copy p [print length? s] ] close p --8<----- %n1.gif --------------------- #{ 47494638396125002500B30000FFFFFF0000008A8A8A454545111111EFEFEF55 55559A9A9AAAAAAABBBBBB666666DFDFDF212121CFCFCF31313176767621F904 00000000002C000000002500250000048B10C849ABBD38EBCDBBFF5C21086089 104190945D61A4A9C16E0D03A7C39C1DE81DE43ACBC30753042D30870F71AC04 1808C18DD1BC2C7C87AA45712314B494868F049E0CBADF3220E133AA01AFDBE2 7DBD01D5DC1BF3DD4B51DF07636F703E737C37326F6C37596F5237696A713883 006730646F97292B83083799803DA16F22029D95A9AAABACADAE0011003B } --8<----- %n2.gif --------------------- #{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} --8<----- %n3.gif --------------------- #{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} --- Website: http://www.codeconscious.com Rebsite: http://www.codeconscious.com/index.r

At first, I was trying to get around reconstituting the gif's, but I created an error condition. I finally faithfully recreated your set-up (except for paths and names), *shutdown and restarted* Apache, and then it worked ... at least for the REBOL HTTP retrieval. (With my testing, I have begun to wonder if the Windows Apache doesn't have a buglet with CGI, because I've noticed this behavior before; just not reproducibly. ???) Now just need that ActiveX component for IE6.0 .... --Scott Jones

enjoy a

I agree with you, but it Tim's call he has the need. For me the excercise has been informative. :^) Regards, Brett.

Well I may go down in flames for this, but what the hell.. How about using Flash as a cross platform player? FlashMX can dynamically load JPEG images now directly at runtime. For Webcam, polling still makes sense, but if you really want 'push' open an XMLSocket in Flash to talk with the server. Although called XMLSocket, it does not have to be actual XML streams. It just means the connection stays open until closed. Oldes' wonderful %make-swf.r can also be used a potent server-side tool in this scheme. It can generate SWFs from images, as well as all sort so tricks to embed Actionscript, buttons, PNGs, localConnection object, dynamic timestamps, metadata, whatever.. http://oldes.multimedia.cz/swf/ LocalConnection is incredible. Makes it trivial to have real-time event passing between people using flash movies, as long as they are on the same server. That means you can be sure people are looking at the same thing at the same time.. browsing or updating image sequences etc. Once you opt for an ActiveX inclusive solution you may as well really enjoy a modern Flash format for all its virtues. ./Jason

Apparently, I did not read your email closely enough. Yes, with REBOL, I did not get 3 "packages" (my term), but one lump payment. I have no clue what IE does with the info, but I do recall it took about 3 seconds to finish. When I set the wait to 3 seconds, it took over 9 seconds for the 3 packages, so I *think* Apache is doing OK in that regard. BTW, and possibly of no particular meaning whatsoever, yesterday I did a search of registry, and found an entry that IE does not cache the multipart/mixed and multipart/x-mixed-replace, I guess because the later could be huge files. I thought it was interesting that it was there at all!

Also, btw, I my Apache did not offer the multipart/x-mixed-replace as a mime type, so I added it to mime.types in the conf folder. ...

... OK, I misunderstood. Maybe I misread, which I seem to be doing *a lot* these days. Time for a vacation/holiday. Finally, thanks, Brett. Your code and effort were appreciated by me. I have no use at this time for this technique, but it is nice to keep in practice with CGI stuff. --Scott Jones

GSJ> Me too! ;-) I think Tim is trying to send a sequence of jpeg images to the browser, so that it looks like a video. (I wonder, wouldn't sending a video be simpler?) GSJ> huge). It is important to note that the "x" in the mime name means GSJ> experimental. I guess is it too experimental for Microsoft's more recent GSJ> browsers to fully and reliably support. Probably, but it has been around since HTTP 1.0 specs if I recall correctly, and probably was in the original MIME specs too. It's something quite old, but probably since very few ever used it MS decided it was not important. (Or maybe they have their own incompatible way of doing the same thing...) Regards, Gabriele. -- Gabriele Santilli <[g--santilli--tiscalinet--it]> -- REBOL Programmer Amigan -- AGI L'Aquila -- REB: http://web.tiscali.it/rebol/index.r

browser until it quits. I've tried setting the mode of the output port to no-wait, but no change. Can someone solve this? I believe that web servers collect all the output of a CGI program, parse the output, and then send it to the browser/internet. There might be a setting in the web server software, to not parse the output of CGI programs? I hope that helps! Andrew Martin ICQ: 26227169 http://valley.150m.com/

From: "Andrew Martin"

programs? Hi, Andrew, Yes, you seem to be correct. Earlier, I looked at documentation about this "problem," and it was not to be a particular problem with earlier versions of Apache. On a lark, I decided to try the proposed fixes, which meant doing two things. Change the name of the script to begin with "nph-" meaning non-parsed-headers (or something like that). Second, because the headers are not parsed, one must then send the appropriate headers (namely, "HTTP/1.1 200). I was successful in using a multipart/x-mixed-replace for a "streamed" text, and IE6 even showed the document! Brett's 3 gif cgi also "showed" up as all three gifs, but the characters, not the pictures. Then when embedded in an html page as an IMG, it downloaded the info, but no pic ever showed up. REBOL as an HTTP client downloaded similar content, but it was again, sent as one final package, not as three parts in a stream. So, at least I could not get it to work much better. I was pleased to at least see the text-only model print anything at all! Thanks for the additional input. --Scott Jones

the

Didn't know about nph - fancy a silly little prefix having such a large impact :^) What I did find out though in the docs for Apache 1.3 (which I'm using) is that the output from CGI is unbuffered anyway. So I gave up with Apache and created a little dedicated REBOL websever based on webserver.r. This little script ignores the request and just outputs three images in sequence just like the CGI was doing. Edit the HTML and make the SRC of IMG to point to http://localhost:8000/ and run the script below in CORE 2.3 (I didn't know how to update it to latest core with the changes to networking - another thing on my learing to do list). What you get in Netscape is a sequence of three images yippee! 1) It would be nice to know how to update this script to the latest Core. 2) This little script proves to me that the content is right for the CGI but not the iteraction between REBOL and Apache for unbuffered output. A mystery. Regards, Brett. REBOL [ Title: "Serve Stream" Date: 10-June-2000 File: %serve-stream.r Purpose: { built on webserver.r to test streaming images. } ] images: [ #{ 47494638396125002500B30000FFFFFF0000008A8A8A454545111111EFEFEF55 55559A9A9AAAAAAABBBBBB666666DFDFDF212121CFCFCF31313176767621F904 00000000002C000000002500250000048B10C849ABBD38EBCDBBFF5C21086089 104190945D61A4A9C16E0D03A7C39C1DE81DE43ACBC30753042D30870F71AC04 1808C18DD1BC2C7C87AA45712314B494868F049E0CBADF3220E133AA01AFDBE2 7DBD01D5DC1BF3DD4B51DF07636F703E737C37326F6C37596F5237696A713883 006730646F97292B83083799803DA16F22029D95A9AAABACADAE0011003B } #{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} #{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} ] listen-port: open/direct/lines tcp://:8000 ; port used for web connections errors: [ 400 "Forbidden" "No permission to access:" 404 "Not Found" "File was not found:" ] send-error: function [err-num file] [err] [ err: find errors err-num insert http-port join "HTTP/1.0 " [ err-num " " err/2 "^/Content-type: text/html^/^/" <HTML> <TITLE> err/2 </TITLE> "<BODY><H1>SERVER-ERROR</H1><P>REBOL Webserver Error:" err/3 " " file newline <P> </BODY> </HTML> ] ] emit: func [data] [ if block? data [data: rejoin data] write-io http-port data length? data write-io http-port CRLF 2 ] boundary-string: "ThisRandomString" boundary: rejoin [CRLF {--} boundary-string] buffer: make string! 1024 ; will auto-expand if needed forever [ http-port: first wait listen-port clear buffer while [not empty? request: first http-port] [ repend buffer [request newline] ] repend buffer ["Address: " http-port/host newline] print mold buffer mime: "text/plain" emit [system/options/cgi/server-protocol { 200 OK}] emit [{Server: } system/options/cgi/server-software] emit [ {Content-Type: multipart/x-mixed-replace;} { boundary-string="} boundary-string {"} ] emit {} emit boundary repeat i 3 [ emit {Content-Type: text/gif} emit {} emit pick images i emit boundary wait 00:00:02 ] emit {Done.} close http-port ]

Hi Guys: I've been preoccupied with another project... I don't think the client wants Flash because they don't want the user to have to install any plug-ins. We got a perl script running some animation, so that I could verify the multipart scheme, so I'll give this another try tomorrow....

-- Tim Johnson <[tim--johnsons-web--com]> http://www.alaska-internet-solutions.com http://www.johnsons-web.com

Correction. :^( A quote: Apache for Windows version 1.3 series is implemented in synchronous calls. This poses an enormous problem for CGI authors, who won't see unbuffered results sent immediately to the browser. This is not the behavior described for CGI in Apache, but it is a side-effect of the Windows port. Apache 2.0 is making progress to implement the expected asynchronous behavior, and we hope to discover that the NT/2000 implementation allows CGI's to behave as documented. Regards, Brett.

Cool!

The mysteries do abound.I was having trouble figuring out whether it was REBOL or Apache, but you have likely said it best, that it is the interaction of the two. I forgot to mention that the nph- issue was written in regard to Apache 0.8!!! I am using 1.3.26, so it seemed like a long shot. There was one other snippet in that advisory that I now cannot recall very well (I've slept since then...), but it said how to have Apache send a raw transmission kind of like as a pass-thru. Of course that was 0.8, and we are up to 1.3, and 2.0 is out there, so who knows whether it would still work. But an untouched pass-through of cgi might work. I guess when Tim get's back to work on this aspect, it might be fun to explore a bit more. Good job, Brett. You get the persistence award for the week. --Scott Jones