World: r3wp

Robert, checksum is not as good as size and date because the file 
has to be downloaded to be sure of the checksum, *or* Ashley has 
to upload an index file containing all the checksums for all the 
distro files every time one of them changes.

I make that statement based on the assumption that it is a regular 
webserver and it doesn't have a way of returning only the checksum 
of a file.

However, it can be a good double check after the decision to download 
has already been made and the file downloaded.

I retract that last statement - it implies you know the checksum 
of the remote file.

Yes, but date/time is a no go. It assumes computer systems are in 
sync, and that time-zones are handled correctly. You know, red-icon 
problem?

anton, what is the problem with checksum? It depends on syncing strategy, 
but you have those files on both side (client, server) anyway. afaik, 
the problem with checksum could be its speed. But since we have open/seek, 
we could do checksumming defined,e.g. 3 parts of files, first xy 
bytes, last xy bytes and defined skip somewhere to the middle of 
the file ... dunno if it would work though :-)

date stamp is completly unreliable, especially with rebol. Red-icons 
problem was in detail described, it is in rambo, I tried several 
times to push for someone to at least look into it, but nothing happened 
....

some kind of checksum plus versioning should do the stuff. But maybe 
we should open Syncing channel for such talks ...

Decided to go with a simple manifest / checksum approach. Long term 
we really need something like Anton outlined previously (multi-level 
caching with the option to control it on an app by app basis). Anyway, 
have a try of this:

	do http://www.dobeash.com/files/rebgui/get-rebgui.r


but be aware that it writes its files to the *current* directory 
(as determined by 'what-dir) and replaces any "base" files with a 
differing checksum (so make a copy of any local changes you have 
made and want to keep).


I've also updated the issues log at: http://www.dobeash.com/it/rebgui/issues.html

lotsa opened issues still there :-)

I do not properly understand the need for auto-updating facility, 
unless provided as general rebgui facility = functionality available 
to custom apps too using rebgui ...

I hope it is so ... otoh do you find Gabriele's update script complicated? 
It has anything programmer needs imo in that regard (judging according 
to my first look, no direct experience though ...)

otoh I prefer the get-rebgui, instead of having separate 0.3.0 plus 
the need to copy over newer stuff :-)

Gabriele's update script ... link?

wait a bit - gabriele put all sorts of his utils here linked few 
days ago ...

http://www.colellachiara.com/soft/Libs

autoupdate.r IIRC ...

there is do-thru and launch-thru, to run things in the right folder.

And a do %tour.r on the end would be perfect :)

A mix could be a self-extracting script for base-distro, and then 
downloading the updates. only a single server-lookup and only a few 
bytes more.

Ashley, great that you have made get-rebgui.r. I'm working on it 
to get it to use the public cache and show progress.

I'd like if you could use checksum/secure to generate the checksums 
in manifest.txt, which is compatible with read-thru, load-thru etc.

Also, if your compressed files begin with the 8 characters "rebpress" 
followed by the compressed data, then we can use the /expand refinement 
of read-thru to auto decompress the data for us.

I've come against a strange problem. When I do this
	read http://www.dobeash.com/files/rebgui/dictionary/American.dat

it stops and gives a network timeout after reading 23% of the file.
bytes: 68432 of total: 293605 = 23.31 %

This is repeatable. I tried with read-thru/progress and a simple 
read.
Bizarre

Seems to be pushing through now. Improved my error reporting on the 
way.

Ashley, totally reworked your get-rebgui.r to use public cache:
http://www.lexicon.net/antonr/rebol/rebgui/get-rebgui.r

Excellent work, implementing all your suggestions now ...

Done.

Works well (do http://www.dobeash.com/files/rebgui/get-rebgui.r) 
except that all the files retain their rebpress prefix! ;) What's 
the best/easiest way to correct that?

already fixed?

Or i am in the wrong folder? looks clean here.

Check the contents of public/www.dobeash.com/files/rebgui

i see. messes the filenames

Maybe a misunderstanding. "rebpress" must be the first part of the 
file-content. not the filename

in read-thru:
        if all [expand find/match data "rebpress"] [

            if error? try [data: decompress skip data 8] [return none]
        ]

Doh! Just reread Anton's message above; 'if your compressed files 
begin with the 8 characters "rebpress" followed by the compressed 
data. Fixing now ...

Is the big spellcheck-data needed?

I found this www.as-u-type.com spell checker works with area fields.

Fixed. Works as intended now.

Spellcheck data is there for the %tour.r demo and is only 287Kb compressed. 
Doesn't change so is only downloaded first time.

As-U-Type is a neat idea except for the fact that:

1) It is Windows only
2) English only
3) Costs money

Do you want Qtask's Spell checker?

Supports multiple languages, is not Windows only, cost nothing.

Not in that order : )

a little self-extractor. kind of rip, but smarter (IMHO). could be 
used to zip the initial download. http://polly.rebol.it/test/test/mixlib/builder/selfextractor.r

Qtask Spell checker ... YES PLEASE. Ideally we could provide links 
to download required functionality / language files as these resources 
may need to be bundled / distributed stand-alone.

Volker: will look at selfextractor.r

Seems still to have a bug, should change filedates. But if you load 
updates with checksum, should not matter anyway.

demo is at http://polly.rebol.it/cgi-bin/../test/test/rebgui/, http://polly.rebol.it/test/test/rebgui/download.r
. the building is in http://polly.rebol.it/test/test/rebgui/build-archive.r

idea is: load archive once, then load updates the get-rebgui-way.