Re: [Patch 0/5] Support windowsize to benefit tftp/pxe download performance.

["Wu, Jiaxin" <jiaxin.wu@intel.com>]

Hi Laszlo,

> 
> If that's the case, should we consider the three drivers under
> "MdeModulePkg/Universal/Network/" deprecated, and should we abandon
> them
> completely?
> 


I think the answer is NO (At least for now). In my opinion, the drivers in MdeModulePkg is also useful in some case, because it already has been used in some platform since it's lightweight to reduce the platform size requirement. As you said below, there is no openssl dependency in MdeModulePkg/ISCSI. 



> If that's the case, then it is really important to document.
> 
> Because, if a user reports a networking-related error (after building
> OVMF without NETWORK_IP6_ENABLE), then I wouldn't like to start
> investigating, just to find out that the issue was fixed in NetworkPkg a
> year earlier.
> 
> Furthermore, if the MdeModulePkg/Universal/Network/ drivers are
> deprecated, when do we intend to actually remove them from the tree?
> 
> Oh, wait, is this related to OpenSSL? When including IScsiDxe from
> NetworkPkg, OpenSSL is required. When including IScsiDxe from
> MdeModulePkg, OpenSSL is not required -- but the networking
> functionality may have some known bugs (which are already fixed in
> NetworkPkg).
> 
> Is this the real trade-off?
> 

For the missed fixes, that's because we didn't take a look whether it also existed in MdeModulePkg, you know, some issues need to be analyzed case by case and most of them are usage related, but actually, it's not so such critical to the other platform. That's why we prefer/try to stay the same of those drivers in MdeModulePkg, if so, it also can help us reduce the development/testing effort against two sets of stacks. What we prefer is to fix the problem / include new feature separately. As I said before, depends on the request/report from customer. So, that's the reason why we recommend to use the stacks in NetworkPkg. Of Couse, if any issue was exposed in MdeModulePkg, we will help to investigate and fix it.

I agree we need document something to highlight that. Siyuan/Ting, if no objection, I will do that in separate patch instead of mixing with this new feature support.


Thanks
Jiaxin