Re: RO full save and images in CSS
On Sat, Jan 29, 2011 at 03:42:54AM +, Harriet Bazley wrote: > I don't think that's actually possible: WebsterXL used to try to do > just this, and the result was that you got all sorts of directories with > random image files in appearing *above* the 'full save' you thought > you'd created. Unfortunately not all elements of a web page are > necessarily loaded from levels subsidiary to the actual HTML file. It's worse than that; any given URL might return different data on each request. (Think about what happens if a "counter image" CGI is included twice.) B.
Re: RO full save and images in CSS
On Sun, Jan 30, 2011 at 12:37:38PM +, Martin Bazley wrote: > The following bytes were arranged on 29 Jan 2011 by Harriet Bazley : > > > Unfortunately not all elements of a web page are necessarily loaded > > from levels subsidiary to the actual HTML file. > > Well, yeah, duh. > > The structure I'd find useful would be something like this: > > !Appname > !Appname.www/example/org/uk > !Appname.www/example/org/uk.index/html > !Appname.www/example/org/uk.images.pic1a/gif > !Appname.www/example/org/uk.thumbs.pic1a/gif > !Appname.www/another/site/com.public.pics.big.screen/jpg > !Appname.!Run (Filer_Run .www/example/org/uk.index/html) The reason that it is not this is to maintain compability with file systems that only support 10 characters. We already decided at the hack weekend that we couldn't be bothered to support these systems any more (after all, we don't support anything older than RISC OS 4.02 anyway.) The idea we actually had was to delegate filenaming for this to the front ends, so they can deal with their local OS's own bizarre file system conventions. Under UNIX, and possibly RISC OS, this would probably contain enough information for a human to identify the file's source, as well as a unique ID. > That way, filename conversion would simply be a matter of replacing > "http://"; with "file:", and relative URLs wouldn't need > changing at all. It requires more thought than this, especially given it has to be portable. > All the same, I get the distinct impression that the only way that would > ever happen would be if I implemented it myself. This is likely. > (Which I would, if > NetSurf wasn't written in this mysterious indecipherable language called > 'C'...) It clearly isn't indecipherable, given millions of people are able to read and write it quite happily :) If it were written in BASIC, you'd have something approximately as useful as WebsterXL. ie, not at all. B.
Re: RO full save and images in CSS
The following bytes were arranged on 29 Jan 2011 by Harriet Bazley : > Unfortunately not all elements of a web page are necessarily loaded > from levels subsidiary to the actual HTML file. Well, yeah, duh. The structure I'd find useful would be something like this: !Appname !Appname.www/example/org/uk !Appname.www/example/org/uk.index/html !Appname.www/example/org/uk.images.pic1a/gif !Appname.www/example/org/uk.thumbs.pic1a/gif !Appname.www/another/site/com.public.pics.big.screen/jpg !Appname.!Run (Filer_Run .www/example/org/uk.index/html) That way, filename conversion would simply be a matter of replacing "http://"; with "file:", and relative URLs wouldn't need changing at all. This would have the advantage that relative URLs in places not currently supported (e.g. CSS) could be much more easily manually fixed, by simply downloading the pictures and placing them in the correct directory. I don't find NetSurf's current structure very useful at all - not just because it makes it impossible to find anything, but because it makes it almost impossible to correct it when things go wrong, such as the problem mentioned in this thread! Another benefit of the above structure would be that you could full-save two different web pages on top of the same application. (To this end, an option to automatically convert URLs to the local structure, regardless of whether it meets the criteria for downloading or not, would be helpful.) All the same, I get the distinct impression that the only way that would ever happen would be if I implemented it myself. (Which I would, if NetSurf wasn't written in this mysterious indecipherable language called 'C'...) -- __<^>__ Follow me on Twitter! --> http://twitter.com/swirlythingy / _ _ \ (Or, um, don't. It's a free country and all that.) ( ( |_| ) ) \_> <_/ === Martin Bazley ==
Re: RO full save and images in CSS
On 29 Jan 2011 Harriet Bazley wrote: > On 25 Jan 2011 as I do recall, > Martin Bazley wrote: >> (The full save code really needs rewriting anyway, to organise things in >> a directory structure with original leafnames intact mimicking the >> structure of an actual website, simultaneously making it more >> user-friendly to browse and easier to transcode URLs for >> > [snip] > I don't think that's actually possible: WebsterXL used to try to do > just this, and the result was that you got all sorts of directories with > random image files in appearing *above* the 'full save' you thought > you'd created. Unfortunately not all elements of a web page are > necessarily loaded from levels subsidiary to the actual HTML file. Not only that, but you might have objects with the same leafnames in different directories so you would still need to ensure unique filenames. It might help to add the usual filename extensions even though they're not needed. > I find Netsurf's approach - to rewrite the whole thing into a RISC OS > application structure and enclose an Inventory file listing the original > sources/names of the files - to be much more useful in practice, and > more elegant. Agreed. -- Richard Porterhttp://www.minijem.plus.com/ mailto:r...@minijem.plus.com I don't want a "user experience" - I just want stuff that works.
Re: RO full save and images in CSS
On 25 Jan 2011 as I do recall, Martin Bazley wrote: > (The full save code really needs rewriting anyway, to organise things in > a directory structure with original leafnames intact mimicking the > structure of an actual website, simultaneously making it more > user-friendly to browse and easier to transcode URLs for > [snip] I don't think that's actually possible: WebsterXL used to try to do just this, and the result was that you got all sorts of directories with random image files in appearing *above* the 'full save' you thought you'd created. Unfortunately not all elements of a web page are necessarily loaded from levels subsidiary to the actual HTML file. I find Netsurf's approach - to rewrite the whole thing into a RISC OS application structure and enclose an Inventory file listing the original sources/names of the files - to be much more useful in practice, and more elegant. -- Harriet Bazley == Loyaulte me lie == The nice thing about standards is that there are so many to choose from
Re: RO full save and images in CSS
The following bytes were arranged on 27 Jan 2011 by Erving : > From: Martin Bazley > Date: 25 Jan 2011 > > > For a good example, try full-saving http://www.beano.com/ . > > I just repeatedly get 'Connection time-out' (Risc PC, RISC OS 4.02 > Netsurf 2.6) though Firefox on an ancient widows laptop displays the > page in seconds. Yes, it does that sometimes; you just have to keep refreshing until it works. -- __<^>__ Follow me on Twitter! --> http://twitter.com/swirlythingy / _ _ \ (Or, um, don't. It's a free country and all that.) ( ( |_| ) ) \_> <_/ === Martin Bazley ==
Re: RO full save and images in CSS
From: Martin Bazley Date: 25 Jan 2011 snip > > For a good example, try full-saving http://www.beano.com/ . > snip I just repeatedly get 'Connection time-out' (Risc PC, RISC OS 4.02 Netsurf 2.6) though Firefox on an ancient widows laptop displays the page in seconds. -- Erving
Re: RO full save and images in CSS
On 25 Jan 2011 Martin Bazley wrote: > I see from the progress page that this is a known issue, but I just want > to request its implementation. > The RISC OS full save, while saving all images referred to in HTML, > doesn't pay any attention to images referred to in CSS, enclosed in > url() brackets. They are neither downloaded nor renamed. I thought that a full save did in fact save all the objects that were downloaded (i.e. before NS ran out of memory, if it did) but that those referenced in stylesheets weren't accessible by clicking Menu on them. Richard -- Richard Porterhttp://www.minijem.plus.com/ mailto:r...@minijem.plus.com I don't want a "user experience" - I just want stuff that works.
Re: RO full save and images in CSS
On Tue, Jan 25, 2011 at 03:00:02PM +, Martin Bazley wrote: > > (The full save code really needs rewriting anyway, to organise things in > a directory structure with original leafnames intact mimicking the > structure of an actual website, simultaneously making it more > user-friendly to browse and easier to transcode URLs for, but that won't > happen in the near future and this is a more important stop-gap.) The need to rewrite the full save functionality was discussed at our last hack weekend. Apart from the issue you describe, the other obvious one is that it is very RISC OS-specific. The only blocker on doing this is nobody wanting to spend the time to do it. B.