I think this is a real problem, expecially in the case of wifi cards. It is sometimes difficult to find out the device name. The h-node database lists the devices, not the chipsets. Different revisions of the same device are listed as different devices (for example WG111v2 and WG111v3). Perhaps we could add the chipset entry so the user can insert the device name and the chipset name.

and help contributors find the names of the device they wish to add in h-node

This is a big problem, we tried to suggest something but perhaps not in the best of ways. Have you got any suggestions? What should be used in place of the lspci command?

There are several places I can suggest, where one should search for the device model's name:

1. On the device itself.
2. On the box that comes with the device.
3. In the manual that comes with the device.
4. In the warranty for the device.
5. If the computer was bought as one whole (usually for portable computers):
• In the manual for the computer
• In the specification of the model in the manufacture's website
• In the website from where the computer was bought from.
6. As last resort, one can use the VendorID:ProductID to search the Web for the device model:
• In Wikipedia there is a very useful article from Intel GPU. There's a table with device models, their corresponding ProductID, and other useful technical information: https://secure.wikimedia.org/wikipedia/en/wiki/Intel_GMA
• Using a search engine one may find information about other device too.

The current form for adding a device tells one should write lspci output (the chipset) in the “model name” field, and the device name in the “other names” field. I think this wrong. I agree with your suggestion, that there should two field – one for the device model and another for the chipset model. The device model name should be the main name of the device, not the chipset.

The lspci command is still useful, especially for finding the chipset name. The list for places where one can search for the device model can be used along with lspci, not instead of.

I'm mostly familiar with GPUs, and I want to rename all of the current “video cards” with their proper names. I think it's best to put all the current lspci names in the “other names” field, and write the device model name in the “model name” field, because this filed it's displayed as the main name of the device.

I'll wait for your comments before I do that.

My suggestions concern mainly the Wi-Fi and Video cards categories. I don't have an opinion for the other categories.

We have deviced to use vendorid:productid to catalogue devices, but that code is the chipset code, not the device code (the device uses some subcodes). So we have decided to list chipsets in place of devices (we have taken this decision after this problem occurred). The device is still listed inside the "other names" field.

Listing the device name (the one written on the box) in the "model name" entry is not possible since there are devices with the same vendorid:productid but different name (see for example this one). You should split the page and create two different pages, one page per device (because the device name, that is the "other name", is different, while the lspci name is the same). But this is not allowed by the system that checks that the vendorid:productid is not present twice in the database.

The "other names" entry is used as well as the "model name" entry during each research (for example in the search page). Also, the "other name" entry is written in the page that lists all the devices (for example http://www.h-node.com/videocards/catalogue/en). So the user can use both (lspci name and device name) during his research.

So please do not replace the "model name" with the "other names", thanks!

Aren't you the who wrote the system? If the system doesn't allow two devices with the same VendorID:ProductID, you should be able to remove that restriction. If you think that the system is better as is, that's different.

I'm not going to push my way much more, but I don't like the way the system is set-up currently. Calling the chipset “model name” is wrong and confusing. Having along an optional “other names” field for the actual device names currently results and will continue resulting in people not specifying the actual name of the device, which is the most useful information when one is looking to buy a peace of hardware. I already explained why this is the most useful information in my first post – because the device name is the one used in when selling a device, not the chipset name.

I believe that buying a device which work with free software is the most useful thing to do with h-node. Currently there are 23 entries for Intel GPUs, but none of the have any information on what the device is called. Because of this, a person currently can't easily use the database for information on what GPU to buy.

Also, even in cases where the actual device name is written in the “other names” field, the chipset name is still displayed frontmost, although the actual device name is the more useful information.

I also like add, that in my first two posts I explained how lspci -nnk actually lists the chipset name of a device, and how two different devices may have the same chipset. Yet you, after my first two posts, explained those two same things to me, like I'm probably not aware of them. This leaves me the impression you didn't pay much attention to what I wrote.

I think there are two reasons to keep listing chipsets (lspci outputs) and group devices having the same chipset.

- other projects are doing so, for example pciid (pci.ids) and usbid (usb.ids) projects. They mainly divide devices using the chipset codes (not the device subcodes, even if they also write the subcodes). In this way we could keep some level of compatibility between the h-node archive and the mentioned archives.

- We need a way to assure that a device is not already in the database. To do so we use the vendorid:productid code of the chipset so we currently ask the user to also insert (it is mandatory) the chipset codes. If we create a page for each device (and not for each chipset), we should ask the user to insert the codes and the subcodes. Everything would be more complicated (not impossible, only more complicated).

Anyway you are right, the model name is used when someone buys a device (not the chipset name). Perhaps we could obtain the same result by highlighting the device name: we could change the label "other names" to "device name" and the label "model name" to "chipset name". So the user that is seraching for a device will most probably use the device name (not the chipset name). We could also change the order in the submission form and in the page showing the device: the new "device name" entry could be placed before the new "chipset name" entry. So the first information that the user see is the "device name" (and not the chipset name). What do you think?

Your solution sounds good. :) This issue report has been bothering me for several years now. I don't remember why I didn't respond in time. Sorry about that! :)

I think this highlights an underlying problem with the h-node design. The software treats products like motherboards and network cards, and chipsets like RTL8029, as the same class of object.

I think the solution is to change the design of the database so that there are two different kinds of object: a "product" object which is linked to one or more "chipset" objects. That way, someone can list their motherboard and link it to their ethernet chipset, the on-board sound chipset, etc.

As a further enhancement, you could also add a "driver" object linked to one or more chipset objects.

Due to the fact that different products with the same chipset don't necessarily work with the same driver, all object types would need information about their usefulness with free software.

Hello! IMHO, using the vendorid:productid is the best option for unequivocally classifying devices (as "primary key" in the database), but the real name should be searchable. I.e. when one searches for a vendorid:productid, it should see ALL products with that chipset. When one searches for a product's real name (as written on the box) it should see MULTIPLE entries for that, if there are multiple revisions, variants, chipsets, etc. IMHO, the real name itself cannot be used as a primary key in the database, because it's equivocal (subtle, one-letter differences, revisions, etc.). Especially for non-brand, cheap items, there are many cases when one cannot figure out the complete and correct name of the device, even if it has the full documentation at hand, CD, manufacturer's website, etc. My 2 cents... :) Răzvan