flex-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Alex Harui <aha...@adobe.com>
Subject Re: RSLs and signing
Date Sun, 10 Feb 2013 16:13:44 GMT

On 2/10/13 8:00 AM, "Harbs" <gavharbs@gmail.com> wrote:

> This is degenerating into a discussion that should probably move to the user
> list, but...
> TLF is a must, but the whole TLF library shouldn't be more than 180-190 KB.
> Nothing like Datagrid is being used. I have a bunch of RichEditableText
> components that I will probably factor out. That might help... On an aside, it
> seems to me that the text components (even the spark ones) are pretty bloated
> and over-engineered. I'm thinking of making some lightweight components for my
> own purposes. Does anyone else see a need for lightweight versions?
> The only two mx components being used is ColorPicker and Alert. With the
> newest release, I should be able to replace those two. I do have one Canvas
> component in use because I couldn't get the behavior I needed with
> BorderContainer. Maybe I'll revisit that. The odd thing was that switching to
> "Spark only" did not bring up any errors for the Canvas component.
And the first screen of your demo app doesn't seem to use multi-line text,
so I would make sure Label is being used for all text which does not require
TLF so you can load it later.

> On Feb 10, 2013, at 5:47 PM, Nicholas Kwiatkowski wrote:
>> The caching rules of RSLs and SWFs are for the most part, are the same.
>> Depending on your IDE, you may be able to build a dependency tree, which
>> should help you determine where your bulk is coming from.  I find that
>> certain components (mx:DataGrid, for example), add in about 250k just for
>> being there because of all the things it depends on.  Some of the Spark
>> text components are equally as heavy because of their dependence on the TLF
>> stuff.  If you could stop using all of the halo components, you will
>> probably be much better off as well.
>> -Nick
>> On Sun, Feb 10, 2013 at 10:41 AM, Harbs <harbs.lists@gmail.com> wrote:
>>> The numbers were for release.
>>> The debug size using RSLs is about 1 MB.
>>> I'm not really sure if modules can help. There are not many modular
>>> components in the app. Maybe I can load the image browser as a module, but
>>> I don't know how much of a difference that will make. There are a number of
>>> palettes that might be candidates. I'll see what I can do on that front,
>>> but I don't have high hopes. My bigger concern is really the Flex libs
>>> which have more bulk than the whole app... Is there a good way of figuring
>>> out where the bulk is coming from?
>>> If I'm reading you right, the caching of swfs is actually more persistent
>>> than the caching of unsigned RSLs. Right?
>>> On Feb 10, 2013, at 5:19 PM, Alex Harui wrote:
>>>> The only advantage to un-signed RSLs is if you serve more than one SWF
>>> that
>>>> uses them from your domain.  SWFs end up on disk in a browser cache (if
>>>> there is one and within the limitations of that cache) so then there is a
>>>> probability you won't have to download some code.
>>>> Apache Flex will hopefully release often.  The Framework RSLs were
>>>> version-specific, so releasing often further lowers your chances of any
>>>> benefit even if we did have a way to serve cross-domain RSLs.
>>>> I suppose we could try to host RSLs at some known place, but browser
>>> cache
>>>> limitations would still apply, and you'd want a custom domain name
>>> otherwise
>>>> you'd expose yourself to cross-domain scripting.
>>>> Are your SWF size numbers for release mode or debug mode?  Using modules
>>>> carefully can lower the size of the initial load.
>>>> On 2/10/13 6:54 AM, "Harbs" <harbs.lists@gmail.com> wrote:
>>>>> Okay. Like you said this sucks.
>>>>> I'm looking to moving from Flex 4.5 to 4.9 in the next few weeks. I just
>>>>> changed my compile settings to merge instead of using RSLs and the app
>>> went
>>>>> from a little over 600 KB to 1.4 MB. :-(
>>>>> I clearly have a lot of work to do removing dependency on a lot of
>>> classes and
>>>>> getting rid of dependency on mx components (I have a very few in use,
>>> but the
>>>>> ones that I'm using will be hard to replace with Spark.)
>>>>> I'm still not sure why Flash can't cache  third party signed RSLs, but
>>> there's
>>>>> not much to be gained by kvetching about it. I doubt they'll add that
>>> as a
>>>>> feature to FlashŠ
>>>>> Harbs
>>>>> On Feb 10, 2013, at 4:37 PM, Nicholas Kwiatkowski wrote:
>>>>>> When I say signed, I'm meaning signed by Adobe.  There really is
>>>>>> little benefit to sign an RSL with our certificates, as they are
>>> the web
>>>>>> of trust of the Flash Player.
>>>>>> From what I've been told, unless it is signed by Adobe, it is not
>>>>>> the persistent cache, so it is not cached on disk, period.  This
>>>>>> regardless of the domain that it is on.
>>>>>> This came up VERY early on (maybe even at the Tech Summit -- I don't
>>> know,
>>>>>> I wasn't there), and Adobe was pretty straight forward that this
>>> going
>>>>>> to be the case.  Questions came up about having them sign it, but
>>> did
>>>>>> not want to dedicated the resources to do it. Looking back, it would
>>> have
>>>>>> been a pain to have to submit our releases to Adobe for their complete
>>>>>> review before we could do anything -- potentially holding back our
>>> releases
>>>>>> weeks or months.
>>>>>> It was seen as a majority of the Flex work was moving to mobile.
>>> AIR
>>>>>> with mobile, there is no concept of RSLs (everything is embedded
>>> within the
>>>>>> final executable), so it was seen as less of an issue.
>>>>>> -Nick
>>>>>> On Sun, Feb 10, 2013 at 9:27 AM, Harbs <harbs.lists@gmail.com>
>>>>>>> Bah! So they're totally useless.
>>>>>>> swfs are also cached by the browser for that session. Correct?
>>>>>>> Is there any logic to not caching RSLs for the domain that loaded
>>> them?
>>>>>>>> Only signed RSLs are cached on disk.
>>>>>>> Signed meaning signed by Adobe. Right? There's no way to sign
>>> with
>>>>>>> an SSL or code signing certificate. Is there?
>>>>>>> On Feb 10, 2013, at 4:19 PM, Nicholas Kwiatkowski wrote:
>>>>>>>> They are downloaded once per domain, per session.  If you
>>> domain
>>>>>>>> x.comtwice in a session (as defined by your browser), then
it will
>>>>>>>> stay in
>>>>>>>> memory.  If you close your session (typically by closing
>>> browser),
>>>>>>>> then it will be cleared from memory.
>>>>>>>> Only signed RSLs are cached on disk.
>>>>>>>> -Nick
>>>>>>>> On Sun, Feb 10, 2013 at 9:01 AM, Harbs <harbs.lists@gmail.com>
>>> wrote:
>>>>>>>>> I apparently missed this. Yes. It does suck. Are RSLs
reloaded every
>>>>>>> time
>>>>>>>>> for a specific domain, or is it just a cross-domain issue?
>>>>>>>>> If I use RSLs for Flex 4.9 and I update my main app,
do the RSLs get
>>>>>>>>> downloaded every time, or will the RSLs from my domain
be reused? Is
>>>>>>> there
>>>>>>>>> any point in using RSLs at all?
>>>>>>>>> On Feb 10, 2013, at 3:56 PM, Nicholas Kwiatkowski wrote:
>>>>>>>>>> Adobe has (had?) a pretty good explanation on their
>>> Whitepaper.
>>>>>>> It
>>>>>>>>>> boils down to this :
>>>>>>>>>> - They are no longer in control of Flex
>>>>>>>>>> - They are no longer doing security reviews of the
source code
>>>>>>>>>> - They have to sign the Flex package with their security
>>> certificate in
>>>>>>>>>> order for it to be stored in the Flash RSL Cache
>>>>>>>>>> - They won't sign it anymore because they would be
responsible for
>>> any
>>>>>>>>>> security issues that may come out of it.
>>>>>>>>>> Yes, it sucks, but unfortunately, we have to live
with it.
>>>>>>>>>> -Nick
>>>>>>>>>> On Sun, Feb 10, 2013 at 8:49 AM, christofer.dutz@c-ware.de
>>>>>>>>>> christofer.dutz@c-ware.de> wrote:
>>>>>>>>>>> I have to admit, that I don't quite understand
what the inability
>>> to
>>>>>>>>>>> create signed rsls has to do with the usage of
rsls themselves.
>>>>>>>>>>> The problem is that the Flashplayer is able to
install rsls that
>>> are
>>>>>>>>>>> signed by Adobe. Usually the Adobe FDK rsls were
also available in
>>>>>>>>> signed
>>>>>>>>>>> versions (swz files). These were dynamically
loaded the first time
>>>>>>> they
>>>>>>>>>>> were needed and installed by the Flashplayer.
The second time the
>>> libs
>>>>>>>>> were
>>>>>>>>>>> needed the installed versions were used reducing
the download time
>>>>>>>>>>> dramatically. Now the problem is that Adobe won't
sign Apache
>>> SWCs as
>>>>>>>>> they
>>>>>>>>>>> are no longer in charge of the libs code (Understandable).
>>>>>>>>> Apache a
>>>>>>>>>>> key to be able to also create signed RSLs would
eventually open
>>>>>>> serious
>>>>>>>>>>> security problems because a signed manipulated
swz would be used
>>> by
>>>>>>>>> every
>>>>>>>>>>> other website using the same version of a given
>>>>>>>>>>> Coming back to the RSLs ... The difference between
a signed and an
>>>>>>>>>>> unsigned RSL is just, that the unsigned rsl is
loaded on every
>>> visit
>>>>>>> of
>>>>>>>>> a
>>>>>>>>>>> user. As far as I know there is no other difference.
So I don't
>>> quite
>>>>>>>>>>> understand why the lack of availability of signed
rsls should
>>> have any
>>>>>>>>>>> effect on built applications and the default
linking type.
>>>>>>>>>>> Chris
>>>>>>>>>>> -----Urspr├╝ngliche Nachricht-----
>>>>>>>>>>> Von: Harbs [mailto:harbs.lists@gmail.com]
>>>>>>>>>>> Gesendet: Sonntag, 10. Februar 2013 14:19
>>>>>>>>>>> An: dev@flex.apache.org
>>>>>>>>>>> Betreff: RSLs and signing
>>>>>>>>>>> I did not realize that Apache Flex does not use
RSLs by default.
>>>>>>>>>>> What's the story with signing? Is that an issue
with cross-domain
>>>>>>>>>>> security? Is there any way to get an Apache signature
approved for
>>>>>>>>> Flash?
>>>>>>>>>>> Either way, I'd imagine I'd want RSLs for the
simple reason that
>>>>>>>>> updating
>>>>>>>>>>> apps should result in a smaller download.
>>>>>>>>>>> Harbs
>>>>>>>>>>> On Feb 9, 2013, at 9:00 AM, Alex Harui wrote:
>>>>>>>>>>>> The default setting for Apache Flex is to
not use RSLs because
>>> Adobe
>>>>>>>>>>>> cannot sign the Apache Flex RSLs.  That's
probably why your SWF
>>> is
>>>>>>>>>>> bigger.
>>>>>>>>>>>> On 2/8/13 10:31 PM, "grimmwerks" <grimm@grimmwerks.com>
>>>>>>>>>>>> Hey all - long time listener first time caller.
>>>>>>>>>>>> I've taken a project that was originally
4.6 and I flipped it to
>>>>>>> 4.9;
>>>>>>>>>>>> comparing the same code on two computers
- when I build with
>>> the 4.6
>>>>>>>>>>>> sdk I get a swf of 304k (with all the other
extraneous libraries
>>>>>>> such
>>>>>>>>>>>> as osmf, mx, sparkspins, etc) -- whereas
with 4.9 the main sf is
>>>>>>>>>>>> 1.1mb -- that's a huge difference with no
other changes in code
>>> no?
>>>>>>>>>>>> Garry Schafer
>>>>>>>>>>>> grimmwerks
>>>>>>>>>>>> grimm@grimmwerks.com
>>>>>>>>>>>> portfolio: www.grimmwerks.com/
>>>>>>>>>>>> --
>>>>>>>>>>>> Alex Harui
>>>>>>>>>>>> Flex SDK Team
>>>>>>>>>>>> Adobe Systems, Inc.
>>>>>>>>>>>> http://blogs.adobe.com/aharui
>>>> --
>>>> Alex Harui
>>>> Flex SDK Team
>>>> Adobe Systems, Inc.
>>>> http://blogs.adobe.com/aharui

Alex Harui
Flex SDK Team
Adobe Systems, Inc.

View raw message