AIDA URL resolution patch

> Hello Janko and all,
>
> I've started playing around with AIDA/Web. I was quite impressed that
> it can handle m17n string quite well! I was able to confirm Japanese string
> is displayed and submitted without any changes to AIDA.
>
> However, I found a little problem in multibyte URL resolution.
>
> In HTTPRequest>>uriString, it returns request uri "as is", so it is
> not encoded as Smalltalk internal m17n string. I think it is OK in Swazoo
> level.
>
> But AIDASite>>objectTo:forSession: uses the raw string for resolving object,
> It fails because urlResolver's dictionary's keys are Smalltalk string.
>
> I think it should be converted as Smalltalk internal m17n string.
> Please see the attached file.
>
> P.S.
> I 've also found some methods that should be fixed.
> (Maybe you have noticed them, but just FYI).
>
> AIDASite class>>default. (there are some senders. But method is not
> implemented).
> AIDASite backupToFile. (In Squeak, there is no BOSS)
> AIDASite restoreSiteNamed: (In Squeak, there is no BOSS)
>
> Thanks again for releasing this great web app server.
>
> Cheers,
> ---
> [:masashi | ^umezawa]
>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> Aida mailing list
> [hidden email]
> http://lists.aidaweb.si/mailman/listinfo/aida

> Hi Masashi,
>
> Nice to hear that Japanese works well on Aida! And yes, if we like
> mention that Aida has nice looking Urls, they must be nice in all
> languages! I admit that here we are a bit lazy, in my country for
> instance we just downgrade our writings to ASCII. Due to inertia
> probably, because no long ago that was the only possibility.
>
> Your patch is a right way to go, but maybe I'd put it in URLResolver
> instead. I'll try this by myself and test with our characters.
>
> Few words about encodings in Aida: on the web Aida uses UTF-8 encoding
> only while internally all strings must be Unicode. That means in Squeak
> Byte/WideStrings, on VW Byte/TwoByte/FourByteStrings. Keep in mind that
> we never preserves strings in UTF-8 internally.
>
> But with your patch you actually made Aida with Japanese Urls, did you
> try that already?
>
> Best regards
> JAnko
>
> Masashi UMEZAWA wrote:
> > Hello Janko and all,
> >
> > I've started playing around with AIDA/Web. I was quite impressed that
> > it can handle m17n string quite well! I was able to confirm Japanese string
> > is displayed and submitted without any changes to AIDA.
> >
> > However, I found a little problem in multibyte URL resolution.
> >
> > In HTTPRequest>>uriString, it returns request uri "as is", so it is
> > not encoded as Smalltalk internal m17n string. I think it is OK in Swazoo
> > level.
> >
> > But AIDASite>>objectTo:forSession: uses the raw string for resolving object,
> > It fails because urlResolver's dictionary's keys are Smalltalk string.
> >
> > I think it should be converted as Smalltalk internal m17n string.
> > Please see the attached file.
> >
> > P.S.
> > I 've also found some methods that should be fixed.
> > (Maybe you have noticed them, but just FYI).
> >
> > AIDASite class>>default. (there are some senders. But method is not
> > implemented).
> > AIDASite backupToFile. (In Squeak, there is no BOSS)
> > AIDASite restoreSiteNamed: (In Squeak, there is no BOSS)
> >
> > Thanks again for releasing this great web app server.
> >
> > Cheers,
> > ---
> > [:masashi | ^umezawa]

> Hi Janko,
>
> Thanks for looking into my patch.
> I've tested the patch in Japanese included URLs, and it worked fine in
> both IE and Firefox.
>
> I will try your fix in Japanese URLs and do more tests.
>
> Cheers,
>
> 2008/1/14, Janko Mivšek <[hidden email]>:
>> Hi Masashi,
>>
>> Nice to hear that Japanese works well on Aida! And yes, if we like
>> mention that Aida has nice looking Urls, they must be nice in all
>> languages! I admit that here we are a bit lazy, in my country for
>> instance we just downgrade our writings to ASCII. Due to inertia
>> probably, because no long ago that was the only possibility.
>>
>> Your patch is a right way to go, but maybe I'd put it in URLResolver
>> instead. I'll try this by myself and test with our characters.
>>
>> Few words about encodings in Aida: on the web Aida uses UTF-8 encoding
>> only while internally all strings must be Unicode. That means in Squeak
>> Byte/WideStrings, on VW Byte/TwoByte/FourByteStrings. Keep in mind that
>> we never preserves strings in UTF-8 internally.
>>
>> But with your patch you actually made Aida with Japanese Urls, did you
>> try that already?
>>
>> Best regards
>> JAnko
>>
>> Masashi UMEZAWA wrote:
>>> Hello Janko and all,
>>>
>>> I've started playing around with AIDA/Web. I was quite impressed that
>>> it can handle m17n string quite well! I was able to confirm Japanese string
>>> is displayed and submitted without any changes to AIDA.
>>>
>>> However, I found a little problem in multibyte URL resolution.
>>>
>>> In HTTPRequest>>uriString, it returns request uri "as is", so it is
>>> not encoded as Smalltalk internal m17n string. I think it is OK in Swazoo
>>> level.
>>>
>>> But AIDASite>>objectTo:forSession: uses the raw string for resolving object,
>>> It fails because urlResolver's dictionary's keys are Smalltalk string.
>>>
>>> I think it should be converted as Smalltalk internal m17n string.
>>> Please see the attached file.
>>>
>>> P.S.
>>> I 've also found some methods that should be fixed.
>>> (Maybe you have noticed them, but just FYI).
>>>
>>> AIDASite class>>default. (there are some senders. But method is not
>>> implemented).
>>> AIDASite backupToFile. (In Squeak, there is no BOSS)
>>> AIDASite restoreSiteNamed: (In Squeak, there is no BOSS)
>>>
>>> Thanks again for releasing this great web app server.
>>>
>>> Cheers,
>>> ---
>>> [:masashi | ^umezawa]
>

> It is interesting for me that Urls with non-ASCII letters just work,
> without any special treatment. Isn't there some special standard for
> that, IDNA?
>
> Here it is: http://www.ietf.org/rfc/rfc3490
> and http://en.wikipedia.org/wiki/Internationalized_domain_name
>
>
> It seems that IDNA is for domain names only, while Urls are free to be
> encoded UTF8? Masashi, if you have time, would you look at this question
> a bit further?

> According to http://www.w3.org/International/O-URL-code.html:
>
> For worldwide interoperability, URIs have to be encoded uniformly. To
> map the wide range of characters used worldwide into the 60 or so
> allowed characters in a URI, a two-step process is used:
>
>     * Convert the character string into a sequence of bytes using the
>       UTF-8 encoding
>     * Convert each byte that is not an ASCII letter or digit to %HH,
>       where HH is the hexadecimal value of the byte

>> According to http://www.w3.org/International/O-URL-code.html:
>>
>> For worldwide interoperability, URIs have to be encoded uniformly. To
>> map the wide range of characters used worldwide into the 60 or so
>> allowed characters in a URI, a two-step process is used:
>>
>>     * Convert the character string into a sequence of bytes using the
>>       UTF-8 encoding
>>     * Convert each byte that is not an ASCII letter or digit to %HH,
>>       where HH is the hexadecimal value of the byte
>
> Yes. This type of encoding is automatically done via web browser (IE
> and Firefox at least).
>
> In my investigation:
> 1. In SwazooURI>>fromString: receives URI encoded string (like
> '/%E3%83%87%E3%83%A2.html').
> 2. The string is converted to Smalltalk byte string via
> HTTPString>>decodedHTTPFrom:.
> 3. But the string is only byte string, so it should be converted to
> m17n string by AIDASite level (maybe seeing codePage).

>> As you know the Url resolution in Aida is two way, one from url to
>> domain object, another from object to url. Second one is needed for
>> auto-url generation on web pages with just pointing to domain object
>> (#addLinkTo: anObject). And if domain object implements method
>> #preferedUrl, you can suggest Url here. Obviously you can suggest
>> international one too, but I'm not yet sure if this works. So Masashi
>> please try this too.
>
> Greatly, it just works. I've tested domain object links by "Address
> Book" example.
> (Please see the attached file).
>
> It is quite understandable because AIDASite class >>convertToWeb:on:
> is called for generating href stirng (Smalltalk m17n string is
> converted to UTF-8 encoded string). Since the returned html includes
> content type header, (="text/html; charset=UTF-8"), web browser can
> interpret the the string as UTF-8. When the link is clicked, the
> browser encode the href string to URI encoded one. (And AIDA decode it
> for further processing...)
>
> Cheers,
>
>
> ------------------------------------------------------------------------
>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> Aida mailing list
> [hidden email]
> http://lists.aidaweb.si/mailman/listinfo/aida