Context : Several items on the diagnostics and interpretation of coronal loop observations are under debate .

Aims : In this work , we analyze a well-defined loop system detected in a time-resolved observation in several spectral bands to study how far one can go in characterizing the loop structure and evolution .

Methods : The dataset includes simultaneous sequences of images in the 171 Å , 195 Å and 284 Å filter bands of TRACE , and in one filter of Yohkoh/SXT , with a time coverage of about 2.5 hours , and two rasters taken with SoHO/CDS in twelve relevant lines , forming between \log T \approx 5.4 ( O V 629 Å ) and \log T \approx 6.4 ( Fe XVI 360 Å ) .
The loop is initially best visible in the TRACE 195 Å filter band , with some correspondence with the simultaneous SXT images , and later in the 171 Å filter band , with good correspondence with the CDS raster images in the lines with formation temperature around \log T \approx 6.0 - 6.1 .
We have taken as pixel-by-pixel background the latest TRACE , Yohkoh and CDS images where the loop has faded out .
We examine the loop morphology evolution , the light curves , the TRACE filter ratio distribution and evolution , the images and emission measure from the CDS spectral lines .

Results : Our analysis detects that , after background subtraction , the emission along the loop and its evolution are non-uniform , especially in the 171 Å filter band , and that the TRACE 195/171 filter ratio has a moderately non-uniform distribution along the loop and evolves in time .
Both the light curves and the filter ratio evolution indicate a globally cooling loop .
Relatively hot plasma may be present at the beginning while , during the first CDS raster , the data indicate a rather moderate thermal structuring of the loop .

Conclusions : Our data analysis supports a coherent scenario across the different bands and instruments , points out difficulties in diagnostic methods and puts quantitative basis for detailed forward modeling .