* Organizing ALL device data in linuxtv wiki
@ 2009-11-12 17:31 H. Langos
2009-11-12 19:27 ` Devin Heitmueller
2009-11-13 16:08 ` [linux-dvb] " H. Langos
0 siblings, 2 replies; 9+ messages in thread
From: H. Langos @ 2009-11-12 17:31 UTC (permalink / raw)
To: linux-media; +Cc: linux-dvb
Hi there,
Like most wikis of the linuxtv wiki is plagued with duplicate
and out-of-sync information. It is most apparent for devices,
their features, their hardware, and most of all their status
in regard to linux support.
If you need an example of the mess take a look at
http://www.linuxtv.org/wiki/index.php/TerraTec or
http://www.linuxtv.org/wiki/index.php/DVB-T_PCI_Cards
and try to find devices where the information on the
page is consistent with the information on the device
specific page.
I've collected and concentrated data on about 150 devices
(mostly DVB-T USB) in one place and experimented a lot
with "wiki template voodoo". The result is a kind of minimalistic
database application within the wiki (retaining the advantages
of the wiki, like mature history and undo function,
low entry threshold for new users....)
Having _one_ article that holds the data means that there is
just one place to update and maintain while the information
can be shown in lots of articles (or inside the same article
with different levels of detail).
That article can be thought of as one big table with columns
for attributes and rows for devices.
Selection of columns is done by choosing the appropriate
table templates (there are currently three different ones
for different levels of detail).
Selection of rows is done by passing selection
attributes and selection values.
It is roughly the equivalent of this sql statement:
SELECT a1,a2,a3... FROM datatemplate
[ WHERE s1 LIKE '%v1%' [ AND s2 LIKE '%v2%'
[AND s3 LIKE '%v3%' [AND s4 LIKE '%v4%' ]]]];
Recently I expanded the "database schema" to contain
the host interface(s) and the supported broadcasting standard(s).
This way the same infrastructure can be used to hold information
on anything from PCIe DVB-S2 to ISA NTSC devices and even
fm radio devices.
The current "database schema" is documented in
http://www.linuxtv.org/wiki/index.php/Template:DVB-T_USB_Devices_ListData#Syntax_and_Semantics
It is still geared towards DVB-T USB devices but I am sure
it can be expanded/modified with little effort to support
other types of hardware with the level of detail needed.
What I'd like to get is
A) Some feedback from users and developers in regard to
additional attributes needed or attributes (or
values) that can be merged.
E.g. I don't have a clue
* if satelite receivers have special attributes, or
* if ANALOG-PAL /-NTSC /-SECAM should be listed separately or
* if it would make sense to include links from a device
straight to the linux kernel's driver blob via
kernel.org's gitweb.
Please, keep in mind that the table is not there
to replace device specific pages with all their
detail but definetly should be part of the device's
page.
B) Ideas about how to handle oem devices (clones). My
idea is to include them with just the "vendor" and
"device" attribute (so that users can easily find the
device in the table looking for the vendor) and to
use the "supported" attribute to indicate that this
device is just a clone of some other device. Problem
here is that you can't really link to that other
device directly or use data from that other device.
(Sorry, no sql JOIN operation on tables :-))
Here are some examples of stuff that already works:
Differnet views on same device in one article.
http://www.linuxtv.org/wiki/index.php/MSI_DigiVox_mini_II_V3.0
http://www.linuxtv.org/wiki/index.php/Pinnacle_PCTV_nano_Stick_(73e)
Selection on demodulator af9015
http://www.linuxtv.org/wiki/index.php/Afatech_AF9015
Selection on hostinterface USB and stadard DVB-T
(still a rather messy page as lots of old devices still
need to be merged/moved into the "database")
http://www.linuxtv.org/wiki/index.php/DVB-T_USB_Devices
Some more examples and experiments:
http://www.linuxtv.org/wiki/index.php/HLPlayground2
http://www.linuxtv.org/wiki/index.php/HLPlayground2/doubleselection
I hope the outlook of having just one place to update
will inspire and motivate developers and wiki users alike
to keep information on their device/driver up to date.
cheers
-henrik
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: Organizing ALL device data in linuxtv wiki
2009-11-12 17:31 Organizing ALL device data in linuxtv wiki H. Langos
@ 2009-11-12 19:27 ` Devin Heitmueller
2009-11-12 19:29 ` Devin Heitmueller
2009-11-13 16:08 ` [linux-dvb] " H. Langos
1 sibling, 1 reply; 9+ messages in thread
From: Devin Heitmueller @ 2009-11-12 19:27 UTC (permalink / raw)
To: Linux Media Mailing List
On Thu, Nov 12, 2009 at 12:31 PM, H. Langos <henrik-dvb@prak.org> wrote:
> Hi there,
>
> Like most wikis of the linuxtv wiki is plagued with duplicate
> and out-of-sync information. It is most apparent for devices,
> their features, their hardware, and most of all their status
> in regard to linux support.
>
> If you need an example of the mess take a look at
> http://www.linuxtv.org/wiki/index.php/TerraTec or
> http://www.linuxtv.org/wiki/index.php/DVB-T_PCI_Cards
> and try to find devices where the information on the
> page is consistent with the information on the device
> specific page.
>
> I've collected and concentrated data on about 150 devices
> (mostly DVB-T USB) in one place and experimented a lot
> with "wiki template voodoo". The result is a kind of minimalistic
> database application within the wiki (retaining the advantages
> of the wiki, like mature history and undo function,
> low entry threshold for new users....)
>
> Having _one_ article that holds the data means that there is
> just one place to update and maintain while the information
> can be shown in lots of articles (or inside the same article
> with different levels of detail).
>
> That article can be thought of as one big table with columns
> for attributes and rows for devices.
>
> Selection of columns is done by choosing the appropriate
> table templates (there are currently three different ones
> for different levels of detail).
>
> Selection of rows is done by passing selection
> attributes and selection values.
>
>
> It is roughly the equivalent of this sql statement:
> SELECT a1,a2,a3... FROM datatemplate
> [ WHERE s1 LIKE '%v1%' [ AND s2 LIKE '%v2%'
> [AND s3 LIKE '%v3%' [AND s4 LIKE '%v4%' ]]]];
>
>
> Recently I expanded the "database schema" to contain
> the host interface(s) and the supported broadcasting standard(s).
> This way the same infrastructure can be used to hold information
> on anything from PCIe DVB-S2 to ISA NTSC devices and even
> fm radio devices.
>
> The current "database schema" is documented in
> http://www.linuxtv.org/wiki/index.php/Template:DVB-T_USB_Devices_ListData#Syntax_and_Semantics
>
> It is still geared towards DVB-T USB devices but I am sure
> it can be expanded/modified with little effort to support
> other types of hardware with the level of detail needed.
>
> What I'd like to get is
>
> A) Some feedback from users and developers in regard to
> additional attributes needed or attributes (or
> values) that can be merged.
>
> E.g. I don't have a clue
> * if satelite receivers have special attributes, or
> * if ANALOG-PAL /-NTSC /-SECAM should be listed separately or
> * if it would make sense to include links from a device
> straight to the linux kernel's driver blob via
> kernel.org's gitweb.
>
> Please, keep in mind that the table is not there
> to replace device specific pages with all their
> detail but definetly should be part of the device's
> page.
>
> B) Ideas about how to handle oem devices (clones). My
> idea is to include them with just the "vendor" and
> "device" attribute (so that users can easily find the
> device in the table looking for the vendor) and to
> use the "supported" attribute to indicate that this
> device is just a clone of some other device. Problem
> here is that you can't really link to that other
> device directly or use data from that other device.
> (Sorry, no sql JOIN operation on tables :-))
>
>
> Here are some examples of stuff that already works:
>
> Differnet views on same device in one article.
>
> http://www.linuxtv.org/wiki/index.php/MSI_DigiVox_mini_II_V3.0
> http://www.linuxtv.org/wiki/index.php/Pinnacle_PCTV_nano_Stick_(73e)
>
>
> Selection on demodulator af9015
>
> http://www.linuxtv.org/wiki/index.php/Afatech_AF9015
>
>
> Selection on hostinterface USB and stadard DVB-T
> (still a rather messy page as lots of old devices still
> need to be merged/moved into the "database")
>
> http://www.linuxtv.org/wiki/index.php/DVB-T_USB_Devices
>
>
> Some more examples and experiments:
>
> http://www.linuxtv.org/wiki/index.php/HLPlayground2
> http://www.linuxtv.org/wiki/index.php/HLPlayground2/doubleselection
>
>
> I hope the outlook of having just one place to update
> will inspire and motivate developers and wiki users alike
> to keep information on their device/driver up to date.
>
> cheers
> -henrik
Hello Henrik,
I have to wonder if maybe we are simply using the wrong tool for the
job. Perhaps it would make sense to make a really simple web frontend
to a simple database for devices. At least initially it would only
really need two tables. Something along the lines of the following
VENDOR_TABLE
int vendor_id (index)
text vendor name
PRODUCT_TABLE
int product_id (index)
text product_name
int bus_type (0=usb, 1=pci, 2=pcie)
text usb id
text pci id
int vendor_id (links to vendors table)
text full_article_page_url
bool hardware_supports_dvb_t
bool hardware_supports_dvb_c
bool hardware_supports_atsc
bool hardware_supports_qam256
bool hardware_supports_pal
bool hardware_supports_ntsc
bool hardware_has_mpeg_decoder
bool hardware_has_ir
bool linux_supports_dvb_t
bool linux_supports_dvb_c
bool linux_supports_atsc
bool linux_supports_qam256
bool linux_supports_pal
bool linux_supports_ntsc
bool linux_has_mpeg_decoder
bool linux_has_ir
A simple db frontend like the above would allow users to search on
most of the relevant properties they care about (seeing all devices by
a single manufacturer, looking up devices by USB ID or PCI ID, looking
for devices that support a certain standard, etc)
I feel like the freeform nature of wikis just lends to the information
not being in a structured manner. I don't doubt that a wiki can be
mangled to do something like this, but a real database seems like such
a cleaner alternative.
Devin
--
Devin J. Heitmueller - Kernel Labs
http://www.kernellabs.com
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: Organizing ALL device data in linuxtv wiki
2009-11-12 19:27 ` Devin Heitmueller
@ 2009-11-12 19:29 ` Devin Heitmueller
0 siblings, 0 replies; 9+ messages in thread
From: Devin Heitmueller @ 2009-11-12 19:29 UTC (permalink / raw)
To: Linux Media Mailing List
Just a quick afterthought - bear in mind the schema I proposed is
something I only spent about two minutes on. It would almost
certainly need some more tweaking/cleanup etc. It meant to
communicate a concept, so don't get too tied up in the details of the
exact implementation.
Devin
--
Devin J. Heitmueller - Kernel Labs
http://www.kernellabs.com
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [linux-dvb] Organizing ALL device data in linuxtv wiki
2009-11-12 17:31 Organizing ALL device data in linuxtv wiki H. Langos
2009-11-12 19:27 ` Devin Heitmueller
@ 2009-11-13 16:08 ` H. Langos
2009-11-13 16:38 ` Jan Hoogenraad
1 sibling, 1 reply; 9+ messages in thread
From: H. Langos @ 2009-11-13 16:08 UTC (permalink / raw)
To: linux-media, linux-dvb
Hi Devin,
I'm sorry. I just realized that I was only subscribed to linux-dvb but
not to linux-media. I fixed that now but my reply to your emails will
not have the correct In-Reply-To/References headers.
> I have to wonder if maybe we are simply using the wrong tool for the
> job. Perhaps it would make sense to make a really simple web frontend
> to a simple database for devices. At least initially it would only
> really need two tables. Something along the lines of the following
...
> A simple db frontend like the above would allow users to search on
> most of the relevant properties they care about (seeing all devices by
> a single manufacturer, looking up devices by USB ID or PCI ID, looking
> for devices that support a certain standard, etc)
I've spent some time discussing the pro and contra of an external database
versus a wiki based approach with some of the other wiki admins:
http://www.linuxtv.org/wiki/index.php/User_talk:Hlangos#Further_ramblings...
The most important point there I guess is, that writing a database app is
a piece of cake and a rather nice way of brushing up on one's SQL foo,
but keeping it structure-wise updated for years to come is hard and
boring work.
Also you have to keep in mind that your database app would need to have
at leasts: revision control, undo, user administration.
I'll not go into details but opening such an application to the public
would need a good amount of hard work and not to forget, security reviews.
Stuff that the wiki already has, and (most important) somebody else is
doing that boring maintenance work so that we can concentrate on the
content.
(I know that user administration could be "borrowed" from the mediawiki
but interfacing those applications will mean that you have to keep updating
your code as the mediawiki code evolves.)
> I feel like the freeform nature of wikis just lends to the information
> not being in a structured manner
True, true.
> I don't doubt that a wiki can be mangled to do something like this,
Well. I had some doubts in the begining. :-)
> but a real database seems like such a cleaner alternative.
Cleaner, yes. But I'd rather have it dirty and full of information
than clean, static and empty. (Oh no .. there comes the bazaar
and cathedral metaphor again ... :-) )
The device data is structure wise rather heterogenious. So a relational
database might not be a very efficient way of capturing it.
In my eyes a more valid contendor to the wiki approach would be something
with a document oriented database like couchdb. But still you'd have
to do write all the boring infrastructure stuff like user administration,
history, undo...
TWiki has the ability to rather nicely blend structured data with
unstructured wiki articles. But I thought it more prudent to get
something done with the tools at hand than spend still more time
looking for the perfect tool ;-)
> Just a quick afterthought - bear in mind the schema I proposed is
> something I only spent about two minutes on. It would almost
> certainly need some more tweaking/cleanup etc. It meant to
> communicate a concept, so don't get too tied up in the details of the
> exact implementation.
Jim has collected the attributes he deems important here:
http://www.linuxtv.org/wiki/index.php/User:Jimbley#Semantics
Howeever I see some problems with the envisioned level of detail
regarding linux support when scaled to hundrets of devices:
http://www.linuxtv.org/wiki/index.php/User_talk:Jimbley#Device_Database
We also had a discussion about the different users and the level of
detail they'd need:
http://www.linuxtv.org/wiki/index.php/User_talk:CityK#Help_with_wiki_integration
Two more things:
1.) The wiki approach allows for different "databases" to be maintained
separately (by different people) and still have results shown in one
resulting table.
This could be useful for Vendor pages (listing all devices by that vendor
independent of the boradcasting standard) or for a broadcasting
standard page that lists all e.g. ATSC devices regardless of wether they
have a USB or PCI interface. The only implication of splitting the
databases is that you need to add one line in your "querry" for each
database.
2.) Different devices (regardless of wether they are in the same
"database" or in different ones) can have different sets of attributes.
If you feel that ATSC device should have separate attributes for
"8VSB" and "QAM" you just simply add those attributes to your
devices and write a table template that will display those
attributes (and ignore things like "firmware" or "url")
The only attributes I'd like to have in all devices are "vendor",
"device" and "did" (Device ID).
-henrik
PS: As you see from the number of links to widely different pages,
a wiki is NOT a good solution for discussions. Just to avoid the
impression that wiki's are my "new hammer". :-)
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [linux-dvb] Organizing ALL device data in linuxtv wiki
2009-11-13 16:08 ` [linux-dvb] " H. Langos
@ 2009-11-13 16:38 ` Jan Hoogenraad
2009-11-13 16:45 ` Devin Heitmueller
2009-11-19 16:48 ` H. Langos
0 siblings, 2 replies; 9+ messages in thread
From: Jan Hoogenraad @ 2009-11-13 16:38 UTC (permalink / raw)
To: linux-media, linux-dvb
Would it be possible to store this information in the CODE archives, and
extract it from there ?
Right now, I end up putting essentially the same information into
structures in the driver and into documentation.
This is hard to keep synchronised.
Basic information like device IDs, vendors, demod types, tuners, etc is
already in place in the driver codes.
Getting data from the hg archives (including development branches)
sounds like a cleaner solution.
H. Langos wrote:
> Hi Devin,
>
> I'm sorry. I just realized that I was only subscribed to linux-dvb but
> not to linux-media. I fixed that now but my reply to your emails will
> not have the correct In-Reply-To/References headers.
>
>> I have to wonder if maybe we are simply using the wrong tool for the
>> job. Perhaps it would make sense to make a really simple web frontend
>> to a simple database for devices. At least initially it would only
>> really need two tables. Something along the lines of the following
> ...
>> A simple db frontend like the above would allow users to search on
>> most of the relevant properties they care about (seeing all devices by
>> a single manufacturer, looking up devices by USB ID or PCI ID, looking
>> for devices that support a certain standard, etc)
>
> I've spent some time discussing the pro and contra of an external database
> versus a wiki based approach with some of the other wiki admins:
> http://www.linuxtv.org/wiki/index.php/User_talk:Hlangos#Further_ramblings...
>
> The most important point there I guess is, that writing a database app is
> a piece of cake and a rather nice way of brushing up on one's SQL foo,
> but keeping it structure-wise updated for years to come is hard and
> boring work.
>
> Also you have to keep in mind that your database app would need to have
> at leasts: revision control, undo, user administration.
> I'll not go into details but opening such an application to the public
> would need a good amount of hard work and not to forget, security reviews.
> Stuff that the wiki already has, and (most important) somebody else is
> doing that boring maintenance work so that we can concentrate on the
> content.
>
> (I know that user administration could be "borrowed" from the mediawiki
> but interfacing those applications will mean that you have to keep updating
> your code as the mediawiki code evolves.)
>
>> I feel like the freeform nature of wikis just lends to the information
>> not being in a structured manner
>
> True, true.
>
>> I don't doubt that a wiki can be mangled to do something like this,
>
> Well. I had some doubts in the begining. :-)
>
>> but a real database seems like such a cleaner alternative.
>
> Cleaner, yes. But I'd rather have it dirty and full of information
> than clean, static and empty. (Oh no .. there comes the bazaar
> and cathedral metaphor again ... :-) )
>
> The device data is structure wise rather heterogenious. So a relational
> database might not be a very efficient way of capturing it.
> In my eyes a more valid contendor to the wiki approach would be something
> with a document oriented database like couchdb. But still you'd have
> to do write all the boring infrastructure stuff like user administration,
> history, undo...
>
> TWiki has the ability to rather nicely blend structured data with
> unstructured wiki articles. But I thought it more prudent to get
> something done with the tools at hand than spend still more time
> looking for the perfect tool ;-)
>
>> Just a quick afterthought - bear in mind the schema I proposed is
>> something I only spent about two minutes on. It would almost
>> certainly need some more tweaking/cleanup etc. It meant to
>> communicate a concept, so don't get too tied up in the details of the
>> exact implementation.
>
>
> Jim has collected the attributes he deems important here:
> http://www.linuxtv.org/wiki/index.php/User:Jimbley#Semantics
>
> Howeever I see some problems with the envisioned level of detail
> regarding linux support when scaled to hundrets of devices:
> http://www.linuxtv.org/wiki/index.php/User_talk:Jimbley#Device_Database
>
> We also had a discussion about the different users and the level of
> detail they'd need:
> http://www.linuxtv.org/wiki/index.php/User_talk:CityK#Help_with_wiki_integration
>
>
> Two more things:
>
> 1.) The wiki approach allows for different "databases" to be maintained
> separately (by different people) and still have results shown in one
> resulting table.
>
> This could be useful for Vendor pages (listing all devices by that vendor
> independent of the boradcasting standard) or for a broadcasting
> standard page that lists all e.g. ATSC devices regardless of wether they
> have a USB or PCI interface. The only implication of splitting the
> databases is that you need to add one line in your "querry" for each
> database.
>
>
> 2.) Different devices (regardless of wether they are in the same
> "database" or in different ones) can have different sets of attributes.
>
> If you feel that ATSC device should have separate attributes for
> "8VSB" and "QAM" you just simply add those attributes to your
> devices and write a table template that will display those
> attributes (and ignore things like "firmware" or "url")
>
> The only attributes I'd like to have in all devices are "vendor",
> "device" and "did" (Device ID).
>
> -henrik
>
> PS: As you see from the number of links to widely different pages,
> a wiki is NOT a good solution for discussions. Just to avoid the
> impression that wiki's are my "new hammer". :-)
>
>
> _______________________________________________
> linux-dvb users mailing list
> For V4L/DVB development, please use instead linux-media@vger.kernel.org
> linux-dvb@linuxtv.org
> http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb
>
--
Jan Hoogenraad
Hoogenraad Interface Services
Postbus 2717
3500 GS Utrecht
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [linux-dvb] Organizing ALL device data in linuxtv wiki
2009-11-13 16:38 ` Jan Hoogenraad
@ 2009-11-13 16:45 ` Devin Heitmueller
2009-11-13 18:10 ` H. Langos
2009-11-19 16:48 ` H. Langos
1 sibling, 1 reply; 9+ messages in thread
From: Devin Heitmueller @ 2009-11-13 16:45 UTC (permalink / raw)
To: Jan Hoogenraad; +Cc: linux-media, linux-dvb
On Fri, Nov 13, 2009 at 11:38 AM, Jan Hoogenraad
<jan-conceptronic@hoogenraad.net> wrote:
> Would it be possible to store this information in the CODE archives, and
> extract it from there ?
> Right now, I end up putting essentially the same information into structures
> in the driver and into documentation.
> This is hard to keep synchronised.
>
> Basic information like device IDs, vendors, demod types, tuners, etc is
> already in place in the driver codes.
>
> Getting data from the hg archives (including development branches) sounds
> like a cleaner solution.
The challenge you run into there is that every driver organizes its
table of products differently, and the driver source code does not
expose what features the device supports in any easily easily parsed
manner. Also, it does not indicate what the hardware supports which
is not supported by the Linux driver.
So for example, you can have a hybrid USB device that supports
ATSC/QAM and analog NTSC. The driver won't really tell you these
things, nor will it tell you that the hardware also supports IR but
the Linux driver does not.
It's one of those ideas that sounds reasonable until you look at how
the actual code defines devices.
Devin
--
Devin J. Heitmueller - Kernel Labs
http://www.kernellabs.com
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [linux-dvb] Organizing ALL device data in linuxtv wiki
2009-11-13 16:45 ` Devin Heitmueller
@ 2009-11-13 18:10 ` H. Langos
2009-11-13 20:28 ` Jan Hoogenraad
0 siblings, 1 reply; 9+ messages in thread
From: H. Langos @ 2009-11-13 18:10 UTC (permalink / raw)
To: Devin Heitmueller; +Cc: Jan Hoogenraad, linux-media, linux-dvb
On Fri, Nov 13, 2009 at 11:45:07AM -0500, Devin Heitmueller wrote:
> On Fri, Nov 13, 2009 at 11:38 AM, Jan Hoogenraad
> <jan-conceptronic@hoogenraad.net> wrote:
> > Would it be possible to store this information in the CODE archives, and
> > extract it from there ?
> > Right now, I end up putting essentially the same information into structures
> > in the driver and into documentation.
It would be possible. But that would require the driver developers
to agree on a machine readable format. Either in the code or in the
documentation. (BTW Which documentation would that be? I thought the code
was the only Documentation :-) )
> > This is hard to keep synchronised.
> >
> > Basic information like device IDs, vendors, demod types, tuners, etc is
> > already in place in the driver codes.
> >
> > Getting data from the hg archives (including development branches) sounds
> > like a cleaner solution.
>
> The challenge you run into there is that every driver organizes its
> table of products differently, and the driver source code does not
> expose what features the device supports in any easily easily parsed
> manner. Also, it does not indicate what the hardware supports which
> is not supported by the Linux driver.
>
> So for example, you can have a hybrid USB device that supports
> ATSC/QAM and analog NTSC. The driver won't really tell you these
> things, nor will it tell you that the hardware also supports IR but
> the Linux driver does not.
>
> It's one of those ideas that sounds reasonable until you look at how
> the actual code defines devices.
Yeap. I agree whole heartedly. For some simle drivers you can read that
information from the source. But most drivers support e.g. more than one
tuner and the information which device has which tuner, is not part of
the driver anymore. Rather the driver looks onto the device's i2c bus
to find out which tuner is present. At least this is what I gathered
from browsing through the driver code in order to get my device
table up to date. (see
http://www.linuxtv.org/wiki/index.php/Talk:DVB-T_USB_Devices#Adding_supported_devices_from_kernel_sources
) I don't actually have a clue. So don't take my word for it.
cheers
-henrik
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [linux-dvb] Organizing ALL device data in linuxtv wiki
2009-11-13 18:10 ` H. Langos
@ 2009-11-13 20:28 ` Jan Hoogenraad
0 siblings, 0 replies; 9+ messages in thread
From: Jan Hoogenraad @ 2009-11-13 20:28 UTC (permalink / raw)
To: Devin Heitmueller, Jan Hoogenraad, linux-media
H. Langos wrote:
> On Fri, Nov 13, 2009 at 11:45:07AM -0500, Devin Heitmueller wrote:
>> The challenge you run into there is that every driver organizes its
>> table of products differently, and the driver source code does not
>> expose what features the device supports in any easily easily parsed
>> manner. Also, it does not indicate what the hardware supports which
>> is not supported by the Linux driver.
>>
>> So for example, you can have a hybrid USB device that supports
>> ATSC/QAM and analog NTSC. The driver won't really tell you these
>> things, nor will it tell you that the hardware also supports IR but
>> the Linux driver does not.
>>
>> It's one of those ideas that sounds reasonable until you look at how
>> the actual code defines devices.
>
> Yeap. I agree whole heartedly. For some simle drivers you can read that
> information from the source. But most drivers support e.g. more than one
> tuner and the information which device has which tuner, is not part of
> the driver anymore. Rather the driver looks onto the device's i2c bus
> to find out which tuner is present. At least this is what I gathered
> from browsing through the driver code in order to get my device
> table up to date. (see
> http://www.linuxtv.org/wiki/index.php/Talk:DVB-T_USB_Devices#Adding_supported_devices_from_kernel_sources
> ) I don't actually have a clue. So don't take my word for it.
>
> cheers
> -henrik
I agree, now you spell it out.
I first thought that at least the names of the supported devices should
be readable from the code.
All supported USB IDs can be found easily from
grep -i dvb /lib/modules/`uname -r`/modules.usbmap | sed -e
's/0x0000.*$//' -e 's/ 0x0003/ /' -e 's/^.*://'
As the code lists names for each device, I would expect there would be
an utility like
lsusb -v -d xxxx:xxxx
to list the names from the .devices = sections , just like depmod scans
the device IDs.
However, I have not found such a utility.
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [linux-dvb] Organizing ALL device data in linuxtv wiki
2009-11-13 16:38 ` Jan Hoogenraad
2009-11-13 16:45 ` Devin Heitmueller
@ 2009-11-19 16:48 ` H. Langos
1 sibling, 0 replies; 9+ messages in thread
From: H. Langos @ 2009-11-19 16:48 UTC (permalink / raw)
To: linux-media; +Cc: linux-dvb
Hi Jan,
In order to ease the maintenance work I have started to include direkt
links from the device's entry in the wiki to the source repository's
history of the driver.
E.g. In the entry for Avermedia AverTV A800 you'll
see "Supported in kernel since 2.6.13".
Now the word "kernel" is a hyperlink to the history page of the a800
driver in Linus's kernel repository.
I could change it to point directly at the driver source but my guess is
that the history will be the thing that you'll look at first anyway.
The only additional information one has to enter per device is a parameter
for the {{Supported in Kernel}} template
instead of
{{Supported in Kernel|since=2.6.25}}
you write
{{Supported in Kernel|since=2.6.25|file=drivers/media/dvb/dvb-usb/dib0700_devices.c}}
I could have shortened the path but I wanted to keep it generic.
Does that look usable/helpful? Is history the right landing page?
Is there anything besides Documentation/dvb/cards.txt that I could take
a look at? Currently that file looks rather outdated so reading the
driver source instead seems more promising.
cheers
-henrik
PS: Same can be done for drivers that are maintained in main hg repository and
in branches but I'd like some feedback first.
On Fri, Nov 13, 2009 at 05:38:50PM +0100, Jan Hoogenraad wrote:
> Would it be possible to store this information in the CODE archives, and
> extract it from there ?
> Right now, I end up putting essentially the same information into
> structures in the driver and into documentation.
> This is hard to keep synchronised.
>
> Basic information like device IDs, vendors, demod types, tuners, etc is
> already in place in the driver codes.
>
> Getting data from the hg archives (including development branches)
> sounds like a cleaner solution.
>
> H. Langos wrote:
>> Hi Devin,
>>
>> I'm sorry. I just realized that I was only subscribed to linux-dvb but
>> not to linux-media. I fixed that now but my reply to your emails will
>> not have the correct In-Reply-To/References headers.
>>
>>> I have to wonder if maybe we are simply using the wrong tool for the
>>> job. Perhaps it would make sense to make a really simple web frontend
>>> to a simple database for devices. At least initially it would only
>>> really need two tables. Something along the lines of the following
>> ...
>>> A simple db frontend like the above would allow users to search on
>>> most of the relevant properties they care about (seeing all devices by
>>> a single manufacturer, looking up devices by USB ID or PCI ID, looking
>>> for devices that support a certain standard, etc)
>>
>> I've spent some time discussing the pro and contra of an external database
>> versus a wiki based approach with some of the other wiki admins:
>> http://www.linuxtv.org/wiki/index.php/User_talk:Hlangos#Further_ramblings...
>>
>> The most important point there I guess is, that writing a database app is
>> a piece of cake and a rather nice way of brushing up on one's SQL foo,
>> but keeping it structure-wise updated for years to come is hard and
>> boring work.
>>
>> Also you have to keep in mind that your database app would need to have
>> at leasts: revision control, undo, user administration.
>> I'll not go into details but opening such an application to the public
>> would need a good amount of hard work and not to forget, security
>> reviews.
>> Stuff that the wiki already has, and (most important) somebody else is
>> doing that boring maintenance work so that we can concentrate on the
>> content.
>>
>> (I know that user administration could be "borrowed" from the mediawiki
>> but interfacing those applications will mean that you have to keep updating
>> your code as the mediawiki code evolves.)
>>
>>> I feel like the freeform nature of wikis just lends to the information
>>> not being in a structured manner
>>
>> True, true.
>>
>>> I don't doubt that a wiki can be mangled to do something like this,
>>
>> Well. I had some doubts in the begining. :-)
>>
>>> but a real database seems like such a cleaner alternative.
>>
>> Cleaner, yes. But I'd rather have it dirty and full of information
>> than clean, static and empty. (Oh no .. there comes the bazaar and
>> cathedral metaphor again ... :-) )
>>
>> The device data is structure wise rather heterogenious. So a relational
>> database might not be a very efficient way of capturing it.
>> In my eyes a more valid contendor to the wiki approach would be something
>> with a document oriented database like couchdb. But still you'd have
>> to do write all the boring infrastructure stuff like user administration,
>> history, undo...
>>
>> TWiki has the ability to rather nicely blend structured data with
>> unstructured wiki articles. But I thought it more prudent to get
>> something done with the tools at hand than spend still more time
>> looking for the perfect tool ;-)
>>
>>> Just a quick afterthought - bear in mind the schema I proposed is
>>> something I only spent about two minutes on. It would almost
>>> certainly need some more tweaking/cleanup etc. It meant to
>>> communicate a concept, so don't get too tied up in the details of the
>>> exact implementation.
>>
>>
>> Jim has collected the attributes he deems important here:
>> http://www.linuxtv.org/wiki/index.php/User:Jimbley#Semantics
>>
>> Howeever I see some problems with the envisioned level of detail
>> regarding linux support when scaled to hundrets of devices:
>> http://www.linuxtv.org/wiki/index.php/User_talk:Jimbley#Device_Database
>>
>> We also had a discussion about the different users and the level of
>> detail they'd need:
>> http://www.linuxtv.org/wiki/index.php/User_talk:CityK#Help_with_wiki_integration
>>
>>
>> Two more things:
>>
>> 1.) The wiki approach allows for different "databases" to be maintained
>> separately (by different people) and still have results shown in one
>> resulting table.
>>
>> This could be useful for Vendor pages (listing all devices by that vendor
>> independent of the boradcasting standard) or for a broadcasting
>> standard page that lists all e.g. ATSC devices regardless of wether they
>> have a USB or PCI interface. The only implication of splitting the
>> databases is that you need to add one line in your "querry" for each
>> database.
>>
>>
>> 2.) Different devices (regardless of wether they are in the same
>> "database" or in different ones) can have different sets of attributes.
>>
>> If you feel that ATSC device should have separate attributes for
>> "8VSB" and "QAM" you just simply add those attributes to your
>> devices and write a table template that will display those attributes
>> (and ignore things like "firmware" or "url")
>>
>> The only attributes I'd like to have in all devices are "vendor",
>> "device" and "did" (Device ID).
>>
>> -henrik
>>
>> PS: As you see from the number of links to widely different pages, a
>> wiki is NOT a good solution for discussions. Just to avoid the
>> impression that wiki's are my "new hammer". :-)
>>
>>
>> _______________________________________________
>> linux-dvb users mailing list
>> For V4L/DVB development, please use instead linux-media@vger.kernel.org
>> linux-dvb@linuxtv.org
>> http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb
>>
>
>
> --
> Jan Hoogenraad
> Hoogenraad Interface Services
> Postbus 2717
> 3500 GS Utrecht
>
> _______________________________________________
> linux-dvb users mailing list
> For V4L/DVB development, please use instead linux-media@vger.kernel.org
> linux-dvb@linuxtv.org
> http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb
>
^ permalink raw reply [flat|nested] 9+ messages in thread
end of thread, other threads:[~2009-11-19 16:48 UTC | newest]
Thread overview: 9+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2009-11-12 17:31 Organizing ALL device data in linuxtv wiki H. Langos
2009-11-12 19:27 ` Devin Heitmueller
2009-11-12 19:29 ` Devin Heitmueller
2009-11-13 16:08 ` [linux-dvb] " H. Langos
2009-11-13 16:38 ` Jan Hoogenraad
2009-11-13 16:45 ` Devin Heitmueller
2009-11-13 18:10 ` H. Langos
2009-11-13 20:28 ` Jan Hoogenraad
2009-11-19 16:48 ` H. Langos
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.